MCP servers, over streamable HTTP

A server is a row with a URL; its tools are discovered by a button and
cached, then offered beside the built-in ones. Written by hand rather than
taken from the reference SDK, because that SDK's transport does its own
connecting -- and the one thing that must not be bypassed is check_url on
every hop. Owning the transport is the point; the framing beside it is the
small part.

Sessions are per call: initialize, initialized, the call, a best-effort
DELETE. Caching one wants an owner, a TTL, eviction, a lock and a shutdown
hook, and the server may expire it under all of that anyway -- ToolContext
is a session-free snapshot precisely so nothing in a tool holds live state.

A server's names and descriptions reach the model as instructions and are
bounded before they do; what it returns is escaped preformatted text, never
markdown. Tools are namespaced per server, so two servers exposing "search"
do not collide and neither shadows a built-in.

Also: a round's calls now run together under a semaphore, results indexed
so each tool turn stays paired with its call, and generation.status names
what is running -- a remote tool is latency-bound, and a silent pause is
what a hang looks like.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>

README.md

@@ -51,6 +51,10 @@ runtime. Clone it, `pip install -e .`, run it.
 - **Web search** — offered to the model as a tool it calls when a question needs
   it. DuckDuckGo out of the box (no account, no key), or point it at your own
   SearXNG, or Firecrawl. The sources stay in the transcript
+- **Your own tools** — describe an HTTP call in the admin area (a schema, a URL
+  template, a secret) and a model can make it. Or add an **MCP server** by URL
+  and its tools appear beside the built-in ones. Both restrictable to groups,
+  and neither can be pointed at your own network unless you say so
 - **Speech in and out** — dictate a message and have replies read aloud, against
   any OpenAI-compatible audio endpoint (whisper.cpp, Speaches, Kokoro…). Each
   person picks their own voice
@@ -79,8 +83,8 @@ runtime. Clone it, `pip install -e .`, run it.
 
 **Planned**
 
-Custom tools and MCP servers · agentic execution (local and over SSH) · image
+Agentic execution (local and over SSH) · image generation · OCR for scanned
-generation · OCR for scanned PDFs · semantic search in the library.
+PDFs · semantic search in the library.
 
 See [PLAN.md](PLAN.md) for what is built, what is not, and why.
 

src/lembas/api/admin_tools.py

@@ -31,11 +31,15 @@ from lembas.db.models import (
     SECRET_PLACEMENTS,
     CustomTool,
     Group,
+    McpServer,
 )
 from lembas.services import custom_tools
 from lembas.services import prompts as prompts_service
 from lembas.services import tools as tools_service
 from lembas.services.crypto import UNCHANGED_SENTINEL, decrypt, keep_or_replace, mask
+from lembas.services.fetch import FetchError, check_url
+from lembas.services.mcp import client as mcp_client
+from lembas.services.mcp import registry as mcp_registry
 from lembas.web.templating import render
 
 log = logging.getLogger(__name__)
@@ -413,3 +417,245 @@ async def update_tool(request: Request, db: Db, user: AdminUser, tool_id: str) -
 
     log.info("%s updated custom tool %s", user.email, tool.slug)
     return _back(f"Saved {tool.name}.")
+
+
+# --- MCP servers -------------------------------------------------------------
+MCP_SLUG_PATTERN = re.compile(r"^[a-z0-9][a-z0-9_-]{0,23}$")
+
+
+def _server(db: Db, server_id: str) -> McpServer:
+    server = db.get(McpServer, server_id)
+    if server is None:
+        raise HTTPException(status.HTTP_404_NOT_FOUND, "That server no longer exists.")
+    return server
+
+
+def _mcp_back(message: str = "") -> Response:
+    target = f"/admin/mcp?saved={message}" if message else "/admin/mcp"
+    return RedirectResponse(target, status_code=status.HTTP_303_SEE_OTHER)
+
+
+def _populate_server(server: McpServer, form) -> None:
+    server.name = str(form.get("name") or "").strip()[:120]
+    server.url = str(form.get("url") or "").strip()[:1000]
+    server.guidance = str(form.get("guidance") or "").replace("\r\n", "\n").strip()
+    server.headers_json = _parse_headers(str(form.get("headers") or ""))
+
+    placement = str(form.get("secret_placement") or SECRET_NONE)
+    server.secret_placement = placement if placement in SECRET_PLACEMENTS else SECRET_NONE
+    server.secret_name = str(form.get("secret_name") or "Authorization").strip()[:120]
+
+    server.timeout = _number(
+        form.get("timeout"), default=30, low=mcp_client.MIN_TIMEOUT, high=mcp_client.MAX_TIMEOUT
+    )
+    server.max_chars = _number(
+        form.get("max_chars"), default=8000, low=mcp_client.MIN_CHARS, high=mcp_client.MAX_CHARS
+    )
+    server.position = _number(form.get("position"), default=server.position or 0, low=0, high=999)
+
+    server.allow_private = "allow_private" in form
+    server.enabled = "enabled" in form
+    server.public = "public" in form
+
+    # One checkbox per advertised tool, so an unticked one is absent. The
+    # stored map holds only the refusals; absent means on.
+    if "tool_choices" in form:
+        offered = set(form.getlist("tool_names"))
+        chosen = set(form.getlist("tool_names_on"))
+        server.tool_overrides_json = dict.fromkeys(offered - chosen, False)
+
+
+def _server_problem(db: Db, server: McpServer, form, *, existing_id: str = "") -> str:
+    if not server.name:
+        return "A server needs a name."
+
+    slug = str(form.get("slug") or "").strip().lower()
+    if not MCP_SLUG_PATTERN.match(slug):
+        return (
+            "The identifier must be lowercase letters, digits, hyphens or "
+            "underscores, and at most 24 characters. It prefixes every tool "
+            "name this server offers."
+        )
+    clash = db.scalar(select(McpServer).where(McpServer.slug == slug))
+    if clash is not None and clash.id != existing_id:
+        return f"There is already a server called “{slug}”."
+    server.slug = slug
+
+    try:
+        check_url(server.url, allow_private=True)
+    except FetchError as exc:
+        return exc.message
+    return ""
+
+
+def _server_detail(
+    request: Request, db: Db, server: McpServer, *, is_new: bool, error: str = "", **extra
+):
+    key = f"tool.mcp_{server.slug}" if server.slug else ""
+    overrides = server.tool_overrides_json or {}
+    return render(
+        request,
+        "admin/mcp_detail.html",
+        {
+            "server": server,
+            "is_new": is_new,
+            "error": error,
+            "groups": list(db.scalars(select(Group).order_by(Group.name))),
+            "selected_groups": extra.pop(
+                "selected_groups", {group.id for group in (server.groups if server.id else [])}
+            ),
+            "headers_text": extra.pop("headers_text", _headers_text(server.headers_json)),
+            "tools": [
+                {**entry, "on": overrides.get(entry.get("name"), True)}
+                for entry in (server.tools_json or [])
+                if isinstance(entry, dict)
+            ],
+            "masked": mask(decrypt(server.secret_encrypted)) if server.secret_encrypted else "",
+            "unchanged": UNCHANGED_SENTINEL,
+            "secret_placements": SECRET_LABELS,
+            "prompt_key": key,
+            "prompt_overridden": key in prompts_service.stored(db),
+            **extra,
+        },
+    )
+
+
+@router.get("/admin/mcp")
+async def mcp_page(request: Request, db: Db, user: AdminUser, saved: str = ""):
+    servers = list(db.scalars(select(McpServer).order_by(McpServer.position, McpServer.slug)))
+    return render(
+        request,
+        "admin/mcp.html",
+        {
+            "servers": servers,
+            "counts": {server.id: len(server.tools_json or []) for server in servers},
+            "saved": saved,
+        },
+    )
+
+
+# Registered before /{server_id}, for the reason given above.
+@router.get("/admin/mcp/new")
+async def new_server_page(request: Request, db: Db, user: AdminUser):
+    draft = McpServer(
+        name="",
+        slug="",
+        url="https://",
+        secret_placement=SECRET_NONE,
+        timeout=30,
+        max_chars=8000,
+        enabled=True,
+        public=True,
+        position=0,
+        tools_json=[],
+        tool_overrides_json={},
+    )
+    return _server_detail(request, db, draft, is_new=True)
+
+
+@router.post("/admin/mcp")
+async def create_server(request: Request, db: Db, user: AdminUser) -> Response:
+    form = await request.form()
+    draft = McpServer(headers_json={}, tools_json=[], tool_overrides_json={})
+    _populate_server(draft, form)
+    draft.position = db.scalar(select(func.coalesce(func.max(McpServer.position), -1))) + 1
+
+    problem = _server_problem(db, draft, form)
+    if problem:
+        return _server_detail(
+            request,
+            db,
+            draft,
+            is_new=True,
+            error=problem,
+            headers_text=str(form.get("headers") or ""),
+            selected_groups=set(form.getlist("group_ids")),
+        )
+
+    draft.secret_encrypted = keep_or_replace(str(form.get("secret") or ""), "")
+    draft.groups = _chosen_groups(db, form, public=draft.public)
+    db.add(draft)
+    db.commit()
+
+    # Discovered immediately, the way a new connection's models are: an
+    # administrator who has just typed a URL wants to know whether it answered.
+    count, error = await mcp_registry.refresh(db, draft)
+    log.info("%s added MCP server %s (%d tools)", user.email, draft.slug, count)
+    if error:
+        return _mcp_back(f"Added {draft.name}, but it could not be reached: {error}")
+    return _mcp_back(f"Added {draft.name} — {count} tool(s).")
+
+
+@router.get("/admin/mcp/{server_id}/edit")
+async def edit_server_page(request: Request, db: Db, user: AdminUser, server_id: str):
+    return _server_detail(request, db, _server(db, server_id), is_new=False)
+
+
+@router.post("/admin/mcp/{server_id}/test")
+async def test_server(request: Request, db: Db, user: AdminUser, server_id: str):
+    """Contact the server and cache what it advertises.
+
+    Returns the row fragment, swapped in place, exactly as "Test & refresh"
+    does for a connection.
+    """
+    server = _server(db, server_id)
+    count, error = await mcp_registry.refresh(db, server)
+    message = (
+        f"{server.name}: {error}"
+        if error
+        else f"{server.name}: found {count} tool{'s' if count != 1 else ''}."
+    )
+    return render(
+        request,
+        "admin/_mcp_row.html",
+        {
+            "server": server,
+            "tool_count": len(server.tools_json or []),
+            "message": message,
+            "message_kind": "error" if error else "success",
+        },
+    )
+
+
+@router.post("/admin/mcp/{server_id}/delete")
+async def delete_server(db: Db, user: AdminUser, server_id: str) -> Response:
+    server = _server(db, server_id)
+    name = server.name
+    db.delete(server)
+    db.commit()
+    log.info("%s deleted MCP server %s", user.email, name)
+    return _mcp_back(f"Deleted {name}.")
+
+
+@router.post("/admin/mcp/{server_id}")
+async def update_server(request: Request, db: Db, user: AdminUser, server_id: str) -> Response:
+    server = _server(db, server_id)
+    form = await request.form()
+
+    draft = McpServer(headers_json={}, tools_json=[], position=server.position)
+    _populate_server(draft, form)
+    problem = _server_problem(db, draft, form, existing_id=server.id)
+    if problem:
+        draft.id = server.id
+        draft.secret_encrypted = server.secret_encrypted
+        draft.tools_json = server.tools_json
+        return _server_detail(
+            request,
+            db,
+            draft,
+            is_new=False,
+            error=problem,
+            headers_text=str(form.get("headers") or ""),
+            selected_groups=set(form.getlist("group_ids")),
+        )
+
+    _populate_server(server, form)
+    server.slug = draft.slug
+    server.secret_encrypted = keep_or_replace(
+        str(form.get("secret") or ""), server.secret_encrypted
+    )
+    server.groups = _chosen_groups(db, form, public=server.public)
+    db.commit()
+
+    log.info("%s updated MCP server %s", user.email, server.slug)
+    return _mcp_back(f"Saved {server.name}.")

src/lembas/services/generation.py

@@ -338,13 +338,19 @@ async def _run(generation: Generation) -> None:
                 *payload["messages"],
                 tools_service.assistant_turn(calls, "".join(round_text)),
             ]
-            for call in calls:
+
-                outcome = await tools_service.run_tool(
+            generation.status = _tool_status(calls)
-                    tool_context, call["name"], call["arguments"]
+            generation.touch()
-                )
+            try:
-                generation.tool_events.append(outcome.event)
+                outcomes = await _run_calls(tool_context, calls)
+            finally:
+                generation.status = ""
                 generation.touch()
+
+            for call, outcome in zip(calls, outcomes, strict=True):
+                generation.tool_events.append(outcome.event)
                 messages.append(tools_service.tool_turn(call, outcome.content))
+            generation.touch()
 
             payload = {**payload, "messages": messages}
 
@@ -465,6 +471,49 @@ async def _maybe_compact(generation: Generation) -> None:
         generation.touch()
 
 
+# How many of a round's tool calls may be in flight at once. A bound rather
+# than none: a model that asks for eight would otherwise open eight sockets and
+# eight database sessions at the same moment.
+MAX_PARALLEL_TOOLS = 4
+
+
+def _tool_status(calls: list[dict]) -> str:
+    """What to show while tools run.
+
+    A remote tool -- an HTTP endpoint, an MCP server -- can take seconds with
+    nothing streaming, and a silent pause is exactly what a hang looks like.
+    """
+    if len(calls) == 1:
+        return f"Running {calls[0]['name']}…"
+    return f"Running {len(calls)} tools…"
+
+
+async def _run_calls(context, calls: list[dict]) -> list:
+    """Run one round's calls together, results in call order.
+
+    Sequential was right when every tool was a local database read. A remote one
+    is latency-bound, and three two-second calls in a row are six seconds of a
+    reply looking hung -- while the model has already been told it may ask for
+    several at once.
+
+    The results are indexed rather than appended as they finish, because each
+    tool turn has to line up with the assistant turn's `tool_calls`: an endpoint
+    matching on `tool_call_id` would otherwise pair the right id with the wrong
+    content the moment two calls came back out of order.
+
+    Safe to run together because `run_tool` never raises, so no failure cancels
+    its siblings, and each runner opens its own `session_scope()` against a
+    database in WAL mode with a busy timeout.
+    """
+    limit = asyncio.Semaphore(MAX_PARALLEL_TOOLS)
+
+    async def one(call: dict):
+        async with limit:
+            return await tools_service.run_tool(context, call["name"], call["arguments"])
+
+    return list(await asyncio.gather(*(one(call) for call in calls)))
+
+
 def _pending_text(db, message: Message) -> str:
     """The user turn this reply is answering, for the size estimate."""
     previous = db.scalars(

src/lembas/services/mcp/__init__.py

@@ -0,0 +1,26 @@
+"""An MCP client: remote servers over streamable HTTP.
+
+Three parts. `protocol` is the wire format and nothing else -- it knows no
+database and no HTTP. `client` owns the transport, which is where the SSRF guard
+lives and the reason none of this comes from the reference SDK. `registry` is
+where a row becomes a tool the chat loop can be offered.
+
+Local stdio servers are deliberately absent. Spawning a subprocess is the
+agentic-execution feature, which wants a confirmation model before it does
+anything; a URL is a different act with a different blast radius.
+"""
+
+from lembas.services.mcp.client import McpSpec, call_tool, list_tools, spec_from
+from lembas.services.mcp.protocol import McpError
+from lembas.services.mcp.registry import offer_name, refresh, tool_defs
+
+__all__ = [
+    "McpError",
+    "McpSpec",
+    "call_tool",
+    "list_tools",
+    "offer_name",
+    "refresh",
+    "spec_from",
+    "tool_defs",
+]

src/lembas/services/mcp/client.py

@@ -0,0 +1,273 @@
+"""Talking to a remote MCP server over streamable HTTP.
+
+One session per call, deliberately. A cached session would need an owner, a
+lifetime, eviction, a lock -- a round runs its tools concurrently -- and a
+shutdown hook, and the server may expire it underneath all of that anyway.
+`ToolContext` is a session-free snapshot precisely so that nothing inside a tool
+holds live state. The cost is one extra POST in front of a call that is already
+a network round trip inside a reply taking seconds; the upgrade, if it is ever
+worth it, is a dict in this module and invisible to everything else.
+
+Redirects are followed by hand and every hop is re-checked, for the reason
+`services/fetch.py` gives: an administrator can point this at any URL, and a
+name resolving to 127.0.0.1 walks past any check that only reads the text.
+"""
+
+from __future__ import annotations
+
+import logging
+from collections.abc import AsyncIterator
+from contextlib import asynccontextmanager, suppress
+from dataclasses import dataclass, field
+from typing import Any
+from urllib.parse import quote, urlparse
+
+import httpx
+
+from lembas.db.models import SECRET_BEARER, SECRET_HEADER, SECRET_QUERY, McpServer
+from lembas.services import fetch as fetch_service
+from lembas.services.crypto import decrypt
+from lembas.services.mcp import protocol
+from lembas.services.mcp.protocol import McpError
+
+log = logging.getLogger(__name__)
+
+# A JSON-RPC message is small; anything this large is a server misbehaving, and
+# reading it into memory before parsing is the failure to avoid.
+MAX_MESSAGE_BYTES = 4 * 1024 * 1024
+
+# tools/list is paged. Both bounds exist because the whole list goes into every
+# request as schema.
+MAX_PAGES = 10
+MAX_TOOLS = 100
+
+MIN_TIMEOUT, MAX_TIMEOUT = 1, 120
+MIN_CHARS, MAX_CHARS = 200, 40_000
+
+
+@dataclass(frozen=True)
+class McpSpec:
+    """One server, read while the session was open. See `custom_tools.HttpSpec`."""
+
+    slug: str
+    name: str
+    url: str
+    headers: dict[str, str] = field(default_factory=dict)
+    secret: str = ""
+    secret_placement: str = SECRET_BEARER
+    secret_name: str = "Authorization"
+    timeout: int = 30
+    max_chars: int = 8000
+    allow_private: bool = False
+
+
+def spec_from(row: McpServer) -> McpSpec:
+    return McpSpec(
+        slug=row.slug,
+        name=row.name or row.slug,
+        url=row.url or "",
+        headers=dict(row.headers_json or {}),
+        secret=decrypt(row.secret_encrypted),
+        secret_placement=row.secret_placement,
+        secret_name=row.secret_name or "Authorization",
+        timeout=min(max(int(row.timeout or 0), MIN_TIMEOUT), MAX_TIMEOUT),
+        max_chars=min(max(int(row.max_chars or 0), MIN_CHARS), MAX_CHARS),
+        allow_private=bool(row.allow_private),
+    )
+
+
+def _headers(spec: McpSpec) -> dict[str, str]:
+    headers = {
+        "User-Agent": fetch_service.USER_AGENT,
+        "Content-Type": "application/json",
+        # Both, because a server may answer either for the same request.
+        "Accept": "application/json, text/event-stream",
+        **{str(k): str(v) for k, v in spec.headers.items()},
+    }
+    if spec.secret:
+        if spec.secret_placement == SECRET_BEARER:
+            headers[spec.secret_name or "Authorization"] = f"Bearer {spec.secret}"
+        elif spec.secret_placement == SECRET_HEADER:
+            headers[spec.secret_name or "Authorization"] = spec.secret
+    return headers
+
+
+def _endpoint(spec: McpSpec) -> str:
+    url = fetch_service.check_url(spec.url, allow_private=spec.allow_private)
+    if spec.secret and spec.secret_placement == SECRET_QUERY:
+        joiner = "&" if urlparse(url).query else "?"
+        url = f"{url}{joiner}{quote(spec.secret_name)}={quote(spec.secret, safe='')}"
+    return url
+
+
+class Session:
+    """One initialised conversation with a server."""
+
+    def __init__(self, spec: McpSpec, client: httpx.AsyncClient) -> None:
+        self.spec = spec
+        self._client = client
+        self._url = ""
+        self._headers = _headers(spec)
+        self._session_id = ""
+        self._next_id = 0
+        self.protocol_version = ""
+        self.server_info: dict[str, Any] = {}
+
+    # --- Transport -----------------------------------------------------------
+    async def _post(self, message: dict[str, Any]) -> httpx.Response:
+        current = self._url
+        headers = dict(self._headers)
+        if self._session_id:
+            headers["Mcp-Session-Id"] = self._session_id
+        if self.protocol_version:
+            headers["MCP-Protocol-Version"] = self.protocol_version
+
+        origin = (urlparse(current).scheme, urlparse(current).netloc)
+        for _ in range(fetch_service.MAX_REDIRECTS + 1):
+            try:
+                response = await self._client.post(current, json=message, headers=headers)
+            except httpx.RequestError as exc:
+                raise McpError(f"Could not reach {self.spec.name}: {exc}") from exc
+
+            if not response.is_redirect:
+                return response
+
+            location = response.headers.get("location", "")
+            if not location:
+                raise McpError(f"{self.spec.name} redirected to nowhere.")
+            if response.status_code not in (307, 308):
+                # 301, 302 and 303 turn a POST into a GET, which means nothing
+                # to a JSON-RPC endpoint. Refused rather than guessed at.
+                raise McpError(
+                    f"{self.spec.name} answered {response.status_code}, which would "
+                    "turn the request into a GET. Point the URL at the endpoint itself."
+                )
+
+            current = fetch_service.check_url(
+                str(response.url.join(location)), allow_private=self.spec.allow_private
+            )
+            if (urlparse(current).scheme, urlparse(current).netloc) != origin:
+                headers.pop(self.spec.secret_name or "Authorization", None)
+                origin = (urlparse(current).scheme, urlparse(current).netloc)
+
+        raise McpError(f"{self.spec.name} redirected too many times.")
+
+    def _read(self, response: httpx.Response, *, request_id: int) -> dict[str, Any]:
+        if response.status_code >= 400:
+            raise McpError(f"{self.spec.name} returned HTTP {response.status_code}.")
+        return protocol.result_of(
+            response.content[:MAX_MESSAGE_BYTES],
+            response.headers.get("content-type", ""),
+            request_id=request_id,
+        )
+
+    # --- Lifecycle -----------------------------------------------------------
+    async def open(self) -> None:
+        self._url = _endpoint(self.spec)
+        self._next_id += 1
+        request_id = self._next_id
+
+        response = await self._post(
+            protocol.request(
+                "initialize",
+                {
+                    "protocolVersion": protocol.PROTOCOL_VERSION,
+                    "capabilities": {},
+                    "clientInfo": protocol.CLIENT_INFO,
+                },
+                request_id=request_id,
+            )
+        )
+        # Captured before the body is read: a server that issues one expects it
+        # on everything after this, including the initialized notification.
+        self._session_id = response.headers.get("mcp-session-id", "")
+        result = self._read(response, request_id=request_id)
+
+        self.protocol_version = str(result.get("protocolVersion") or protocol.PROTOCOL_VERSION)
+        info = result.get("serverInfo")
+        self.server_info = info if isinstance(info, dict) else {}
+        if self.protocol_version != protocol.PROTOCOL_VERSION:
+            log.info(
+                "%s speaks MCP %s, we asked for %s",
+                self.spec.name,
+                self.protocol_version,
+                protocol.PROTOCOL_VERSION,
+            )
+
+        await self._post(protocol.notification("notifications/initialized"))
+
+    async def call(self, method: str, params: dict[str, Any] | None = None) -> dict[str, Any]:
+        """One request, re-initialising once if the session has expired."""
+        self._next_id += 1
+        request_id = self._next_id
+        response = await self._post(protocol.request(method, params, request_id=request_id))
+
+        if response.status_code == 404 and self._session_id:
+            # The server dropped the session. One retry, then it is a failure
+            # like any other -- a loop here would be a loop against a server
+            # that has decided to forget us.
+            log.info("%s expired its session; re-initialising", self.spec.name)
+            self._session_id = ""
+            await self.open()
+            self._next_id += 1
+            request_id = self._next_id
+            response = await self._post(protocol.request(method, params, request_id=request_id))
+
+        return self._read(response, request_id=request_id)
+
+    async def close(self) -> None:
+        """Best effort. A server with nothing to clean up answers 405."""
+        if not self._session_id:
+            return
+        headers = {**self._headers, "Mcp-Session-Id": self._session_id}
+        if self.protocol_version:
+            headers["MCP-Protocol-Version"] = self.protocol_version
+        with suppress(httpx.RequestError, McpError):
+            await self._client.delete(self._url, headers=headers)
+
+
+@asynccontextmanager
+async def session_for(spec: McpSpec) -> AsyncIterator[Session]:
+    """An initialised session, closed afterwards whatever happened."""
+    client = httpx.AsyncClient(timeout=spec.timeout, follow_redirects=False)
+    session = Session(spec, client)
+    try:
+        await session.open()
+        yield session
+    finally:
+        await session.close()
+        await client.aclose()
+
+
+async def list_tools(spec: McpSpec) -> tuple[list[dict[str, Any]], dict[str, Any], str]:
+    """Every tool a server advertises, plus what it said about itself.
+
+    Returns (tools, serverInfo, protocolVersion). Bounded at MAX_TOOLS: the list
+    is sent as schema on every request, so a server offering two hundred is a
+    server that would fill the window before anything was asked.
+    """
+    tools: list[dict[str, Any]] = []
+    async with session_for(spec) as session:
+        cursor = ""
+        for _ in range(MAX_PAGES):
+            result = await session.call("tools/list", {"cursor": cursor} if cursor else {})
+            for entry in result.get("tools") or []:
+                cleaned = protocol.clean_tool(entry)
+                if cleaned is None:
+                    log.info("%s advertised an unusable tool entry", spec.name)
+                elif len(tools) < MAX_TOOLS:
+                    tools.append(cleaned)
+            cursor = str(result.get("nextCursor") or "")
+            if not cursor or len(tools) >= MAX_TOOLS:
+                break
+        return tools, session.server_info, session.protocol_version
+
+
+async def call_tool(spec: McpSpec, name: str, arguments: dict[str, Any]) -> tuple[str, bool]:
+    """Run one tool. Returns (text, is_error)."""
+    async with session_for(spec) as session:
+        result = await session.call("tools/call", {"name": name, "arguments": arguments})
+        return protocol.content_to_text(result), bool(result.get("isError"))
+
+
+__all__ = ["McpError", "McpSpec", "Session", "call_tool", "list_tools", "spec_from"]

src/lembas/services/mcp/protocol.py

@@ -0,0 +1,195 @@
+"""The MCP wire format: JSON-RPC 2.0, and what comes back from a tool call.
+
+Written out rather than taken from the reference SDK. The client is a few
+hundred lines of framing, and the SDK's transport does its own connecting --
+which would mean the one thing that must not be bypassed, `fetch.check_url` on
+every hop, being bypassed. Owning the transport is the point; owning the framing
+beside it is the small part.
+
+A response arrives either as one JSON object or as an event stream carrying
+several messages, and a server may choose either for the same request. Both are
+read here so `client.py` does not have to care which it got.
+"""
+
+from __future__ import annotations
+
+import json
+from contextlib import suppress
+from typing import Any
+
+from lembas import __version__
+
+# What we tell a server we speak. A server answering an older version is not
+# refused: several in the wild still answer 2024-11-05 and work perfectly.
+PROTOCOL_VERSION = "2025-06-18"
+
+CLIENT_INFO = {"name": "LLeMbas", "version": __version__}
+
+# A tool's metadata is sent to the model as instructions, so it is bounded here
+# rather than trusted. A server advertising a 40 KB description would spend the
+# context window before the conversation started.
+MAX_DESCRIPTION = 1000
+MAX_SCHEMA_BYTES = 8192
+MAX_NAME = 64
+
+# Content that is not text is described rather than forwarded. A tool turn is a
+# string, images only reach models marked as having vision, and base64 in a tool
+# result is the fastest way to fill a window with nothing.
+UNSUPPORTED = "[{kind}: {detail} — not shown to the model]"
+
+
+class McpError(Exception):
+    """A failed exchange, with a message fit to show an administrator.
+
+    Same contract as `LLMError`, `SearchError` and `FetchError`: the message is
+    the whole error, and is safe to render.
+    """
+
+    def __init__(self, message: str) -> None:
+        super().__init__(message)
+        self.message = message
+
+
+def request(method: str, params: dict[str, Any] | None, *, request_id: int) -> dict[str, Any]:
+    message: dict[str, Any] = {"jsonrpc": "2.0", "id": request_id, "method": method}
+    if params is not None:
+        message["params"] = params
+    return message
+
+
+def notification(method: str, params: dict[str, Any] | None = None) -> dict[str, Any]:
+    message: dict[str, Any] = {"jsonrpc": "2.0", "method": method}
+    if params is not None:
+        message["params"] = params
+    return message
+
+
+def _messages(body: bytes, content_type: str) -> list[dict[str, Any]]:
+    """Every JSON-RPC message in a response body, whichever framing was used."""
+    text = body.decode("utf-8", "replace").strip()
+    if not text:
+        return []
+
+    if "text/event-stream" not in content_type.lower():
+        try:
+            document = json.loads(text)
+        except json.JSONDecodeError as exc:
+            raise McpError(f"That server did not answer with JSON: {exc}") from exc
+        return document if isinstance(document, list) else [document]
+
+    # Frames are `data:` lines gathered until a blank line -- the same framing
+    # services/sse.py writes. Read here rather than shared, because that module
+    # is a writer and this is a reader with a size cap.
+    out: list[dict[str, Any]] = []
+    data: list[str] = []
+    for line in text.splitlines() + [""]:
+        if line.startswith("data:"):
+            data.append(line[5:].lstrip())
+        elif not line.strip() and data:
+            # A frame that is not JSON is a comment or a keep-alive, not a
+            # message; the stream carries both.
+            with suppress(json.JSONDecodeError):
+                out.append(json.loads("\n".join(data)))
+            data = []
+    return out
+
+
+def result_of(body: bytes, content_type: str, *, request_id: int) -> dict[str, Any]:
+    """The result for one request, out of whatever the server sent back.
+
+    Raises `McpError` on a JSON-RPC error member, because that is a failure the
+    administrator needs the words of -- "Unknown tool" and "invalid params" are
+    the two that actually happen.
+    """
+    for message in _messages(body, content_type):
+        if not isinstance(message, dict) or message.get("id") != request_id:
+            continue
+        if "error" in message:
+            error = message["error"] or {}
+            code = error.get("code", "")
+            text = str(error.get("message") or "The server reported an error.")
+            raise McpError(f"{text}{f' (code {code})' if code != '' else ''}")
+        result = message.get("result")
+        return result if isinstance(result, dict) else {}
+
+    raise McpError("That server did not answer the request.")
+
+
+# --- Tool metadata -----------------------------------------------------------
+def clean_tool(entry: Any) -> dict[str, Any] | None:
+    """One advertised tool, bounded. None if there is nothing usable here."""
+    if not isinstance(entry, dict):
+        return None
+    name = str(entry.get("name") or "").strip()
+    if not name or len(name) > MAX_NAME:
+        return None
+
+    schema = entry.get("inputSchema")
+    if not isinstance(schema, dict) or schema.get("type") != "object":
+        schema = {"type": "object", "properties": {}}
+    elif len(json.dumps(schema)) > MAX_SCHEMA_BYTES:
+        # Kept callable rather than dropped: a model can still be told the tool
+        # exists, and an argument it guesses is no worse than not offering it.
+        schema = {"type": "object", "properties": {}}
+
+    return {
+        "name": name,
+        "title": str(entry.get("title") or "")[:MAX_NAME],
+        "description": str(entry.get("description") or "")[:MAX_DESCRIPTION],
+        "schema": schema,
+    }
+
+
+# --- Tool results ------------------------------------------------------------
+def _block_text(block: Any) -> str:
+    if not isinstance(block, dict):
+        return ""
+    kind = str(block.get("type") or "")
+
+    if kind == "text":
+        return str(block.get("text") or "")
+
+    if kind in ("image", "audio"):
+        size = len(str(block.get("data") or ""))
+        detail = f"{block.get('mimeType') or 'unknown type'}, about {size * 3 // 4} bytes"
+        return UNSUPPORTED.format(kind=kind, detail=detail)
+
+    if kind == "resource":
+        resource = block.get("resource")
+        if not isinstance(resource, dict):
+            return ""
+        uri = str(resource.get("uri") or "")
+        if isinstance(resource.get("text"), str):
+            return f"{uri}\n{resource['text']}" if uri else str(resource["text"])
+        detail = f"{uri or 'unnamed'}, {resource.get('mimeType') or 'unknown type'}"
+        return UNSUPPORTED.format(kind="resource", detail=detail)
+
+    if kind == "resource_link":
+        return f"{block.get('name') or 'resource'} ({block.get('uri') or ''})".strip()
+
+    # An addition to the protocol degrades to a note rather than to silence:
+    # a model told nothing came back will say nothing came back.
+    return UNSUPPORTED.format(kind="content", detail=kind or "no type")
+
+
+def content_to_text(result: dict[str, Any]) -> str:
+    """A `tools/call` result as the flat text a tool turn carries."""
+    blocks = result.get("content")
+    parts = [text for block in (blocks or []) if (text := _block_text(block).strip())]
+
+    if not parts and isinstance(result.get("structuredContent"), dict | list):
+        return json.dumps(result["structuredContent"], indent=2, ensure_ascii=False)
+
+    return "\n\n".join(parts)
+
+
+__all__ = [
+    "CLIENT_INFO",
+    "PROTOCOL_VERSION",
+    "McpError",
+    "clean_tool",
+    "content_to_text",
+    "notification",
+    "request",
+    "result_of",
+]

src/lembas/services/mcp/registry.py

@@ -0,0 +1,185 @@
+"""Turning MCP servers into tools a chat can be offered.
+
+Two names per tool. The server has its own, which is what `tools/call` must be
+given; we have an *offered* name, which is what goes in the schema the endpoint
+sees. They differ because two servers both exposing `search` would collide, a
+server exposing `notes_delete` would shadow a built-in, and endpoints accept a
+narrower character set than MCP does. The rename never leaves this module: the
+runner closes over the server's own name.
+
+The advertised list is cached on the row and refreshed by a button, the same
+shape as discovering a connection's models. A server is contacted when an
+administrator asks, not when a chat starts -- a slow server must not be able to
+delay every reply.
+"""
+
+from __future__ import annotations
+
+import logging
+import re
+from datetime import UTC, datetime
+from typing import Any
+
+from sqlalchemy.orm import Session as DBSession
+
+from lembas.db.models import McpServer, User
+from lembas.services import tool_access
+from lembas.services.fetch import FetchError
+from lembas.services.mcp import client
+from lembas.services.mcp.protocol import McpError
+from lembas.services.tools import FAMILY_MCP, ToolContext, ToolDef, ToolOutcome
+
+log = logging.getLogger(__name__)
+
+# What an endpoint will accept as a function name.
+FUNCTION_NAME = re.compile(r"^[a-zA-Z0-9_-]{1,64}$")
+MAX_NAME = 64
+
+# How much of a reply is kept on the message row for the transcript.
+MAX_EVENT_CHARS = 2000
+MAX_SUMMARY_CHARS = 200
+
+
+def offer_name(server_slug: str, tool_name: str, *, taken: set[str]) -> str:
+    """A name the endpoint will accept, unique across everything offered.
+
+    Truncation can collide where the full names would not, so a numeric suffix
+    is appended until it does not. Deterministic given a stable iteration order,
+    which is why rows are walked in (position, slug) order everywhere.
+    """
+    combined = f"{server_slug}_{tool_name}".lower()
+    cleaned = re.sub(r"_+", "_", re.sub(r"[^a-z0-9_-]", "_", combined)).strip("_")
+    candidate = (cleaned or "tool")[:MAX_NAME]
+
+    suffix = 2
+    while candidate in taken:
+        tail = f"_{suffix}"
+        candidate = f"{(cleaned or 'tool')[: MAX_NAME - len(tail)]}{tail}"
+        suffix += 1
+
+    taken.add(candidate)
+    return candidate
+
+
+def _summary(arguments: dict[str, Any]) -> str:
+    parts = [f"{name}={value!r}" for name, value in arguments.items()]
+    text = ", ".join(parts)
+    return text[:MAX_SUMMARY_CHARS]
+
+
+def _runner(spec: client.McpSpec, tool_name: str, offered: str):
+    async def run(context: ToolContext, args: dict[str, Any]) -> ToolOutcome:
+        event = {
+            "name": offered,
+            "kind": "mcp",
+            "label": f"{spec.name} · {tool_name}",
+            "query": _summary(args),
+            "detail": spec.name,
+            "results": [],
+        }
+        try:
+            text, failed = await client.call_tool(spec, tool_name, args)
+        except (McpError, FetchError) as exc:
+            message = exc.message
+            log.info("mcp %s/%s failed: %s", spec.slug, tool_name, message)
+            return ToolOutcome(
+                f"The {tool_name} tool on {spec.name} failed: {message}",
+                {**event, "status": "error", "error": message[:200]},
+            )
+
+        text = text.strip()[: spec.max_chars]
+        if failed:
+            return ToolOutcome(
+                f"The tool reported an error:\n{text}" if text else "The tool reported an error.",
+                {**event, "status": "error", "error": text[:200] or "The tool reported an error."},
+            )
+        if not text:
+            return ToolOutcome(
+                f"{tool_name} returned nothing.", {**event, "status": "ok", "text": ""}
+            )
+        return ToolOutcome(text, {**event, "status": "ok", "text": text[:MAX_EVENT_CHARS]})
+
+    return run
+
+
+def _offered_tools(server: McpServer) -> list[dict[str, Any]]:
+    """The cached entries this server is currently willing to offer."""
+    overrides = server.tool_overrides_json or {}
+    # Absent means on, the rule the model capability flags follow: a tool that
+    # appeared in the last refresh should work rather than silently do nothing.
+    return [
+        entry
+        for entry in (server.tools_json or [])
+        if isinstance(entry, dict) and overrides.get(entry.get("name"), True)
+    ]
+
+
+def tool_defs(
+    db: DBSession, user: User | None, *, everything: bool = False, taken: set[str] | None = None
+) -> list[ToolDef]:
+    """One `ToolDef` per offerable tool across every visible server."""
+    claimed = taken if taken is not None else set()
+    out: list[ToolDef] = []
+
+    for server in tool_access.visible_mcp_servers(db, user, everything=everything):
+        spec = client.spec_from(server)
+        for entry in _offered_tools(server):
+            name = str(entry.get("name") or "")
+            offered = str(entry.get("offer_name") or "") or offer_name(
+                server.slug, name, taken=claimed
+            )
+            claimed.add(offered)
+            if not FUNCTION_NAME.match(offered):
+                continue
+            out.append(
+                ToolDef(
+                    name=offered,
+                    family=f"{FAMILY_MCP}:{server.slug}",
+                    description=entry.get("description") or f"{name}, from {server.name}.",
+                    parameters=entry.get("schema") or {"type": "object", "properties": {}},
+                    run=_runner(spec, name, offered),
+                )
+            )
+
+    return out
+
+
+async def refresh(db: DBSession, server: McpServer) -> tuple[int, str]:
+    """Contact a server and cache what it advertises. Returns (count, error).
+
+    Shaped like `api/admin.py:_refresh_models`, including writing `last_error`
+    and `last_checked_at` on both paths so the row says what happened rather
+    than only whether it worked.
+    """
+    spec = client.spec_from(server)
+    try:
+        advertised, info, version = await client.list_tools(spec)
+    except (McpError, FetchError) as exc:
+        server.last_error = exc.message
+        server.last_checked_at = datetime.now(UTC)
+        db.commit()
+        return 0, exc.message
+
+    taken: set[str] = set()
+    for entry in advertised:
+        entry["offer_name"] = offer_name(server.slug, entry["name"], taken=taken)
+
+    # Choices about tools that are still advertised survive; ones about tools
+    # that have gone are dropped rather than left to accumulate.
+    names = {entry["name"] for entry in advertised}
+    server.tool_overrides_json = {
+        name: on for name, on in (server.tool_overrides_json or {}).items() if name in names
+    }
+
+    server.tools_json = advertised
+    server.server_info = info
+    server.protocol_version = version
+    server.last_error = ""
+    server.last_checked_at = datetime.now(UTC)
+    db.commit()
+
+    log.info("mcp %s advertised %d tool(s)", server.slug, len(advertised))
+    return len(advertised), ""
+
+
+__all__ = ["offer_name", "refresh", "tool_defs"]

src/lembas/services/tools.py

@@ -715,12 +715,19 @@ def _family_allowed(
 def _row_defs(db: DBSession, user: User | None, *, everything: bool = False) -> list[ToolDef]:
     """Tool definitions built from rows, in the order they claim names.
 
-    Imported here rather than at the top because `custom_tools` needs `ToolDef`
+    Custom tools first, then MCP servers, because a custom tool's name is
-    from this module.
+    written by hand and refused if it collides while an MCP tool's is derived
+    and renamed silently -- the one that can adapt should be the one that has to.
+
+    Imported here rather than at the top because both modules need `ToolDef`
+    from this one.
     """
     from lembas.services import custom_tools
+    from lembas.services.mcp import registry as mcp_registry
 
-    return custom_tools.tool_defs(db, user, everything=everything)
+    custom = custom_tools.tool_defs(db, user, everything=everything)
+    taken = {*REGISTRY, *(tool.name for tool in custom)}
+    return [*custom, *mcp_registry.tool_defs(db, user, everything=everything, taken=taken)]
 
 
 def _book(defs: list[ToolDef]) -> dict[str, ToolDef]:
@@ -944,9 +951,11 @@ def _row_source(db: DBSession):
     it -- the override outlives the row.
 
     Gated on the tool's own family, so the guidance appears exactly when the
-    tool it describes is offered and never otherwise.
+    tool it describes is offered and never otherwise. An MCP server gets one
+    fragment rather than one per advertised tool: forty entries on the prompts
+    page is a page nobody would read.
     """
-    from lembas.db.models import CustomTool
+    from lembas.db.models import CustomTool, McpServer
 
     for row in db.scalars(select(CustomTool).order_by(CustomTool.position, CustomTool.slug)):
         yield prompts_service.Fragment(
@@ -959,6 +968,17 @@ def _row_source(db: DBSession):
             default=row.guidance or "",
         )
 
+    for server in db.scalars(select(McpServer).order_by(McpServer.position, McpServer.slug)):
+        yield prompts_service.Fragment(
+            key=f"tool.mcp_{server.slug}",
+            label=server.name or server.slug,
+            group=prompts_service.GROUP_TOOLS,
+            order=600 + server.position,
+            families=(f"{FAMILY_MCP}:{server.slug}",),
+            hint=f"Appears when any tool from {server.name or server.slug} is offered.",
+            default=server.guidance or "",
+        )
+
 
 __all__ = [
     "FAMILIES",

src/lembas/web/templates/admin/_layout.html

@@ -47,6 +47,10 @@
           {{ icon("link", "icon--sm") }}
           <span class="nav-item__label">Tools</span>
         </a>
+        <a class="nav-item {{ 'is-active' if section == 'mcp' }}" href="/admin/mcp">
+          {{ icon("server", "icon--sm") }}
+          <span class="nav-item__label">MCP servers</span>
+        </a>
         <a class="nav-item {{ 'is-active' if section == 'prompts' }}" href="/admin/prompts">
           {{ icon("sparkle", "icon--sm") }}
           <span class="nav-item__label">Prompts</span>

src/lembas/web/templates/admin/_mcp_row.html

@@ -0,0 +1,36 @@
+{% from "_macros.html" import icon %}
+{#
+  One MCP server in the list.
+
+  Swapped in place by the “Test & refresh” button, so this fragment has to be
+  able to render on its own as well as inside the list.
+#}
+<div class="model-row {{ 'is-off' if not server.enabled }}" id="mcp-{{ server.id }}">
+  <div class="model-row__main">
+    <div class="model-row__title">
+      <a class="model-row__name" href="/admin/mcp/{{ server.id }}/edit">{{ server.name }}</a>
+      <span class="badge">{{ tool_count }} tool{{ '' if tool_count == 1 else 's' }}</span>
+      {% if not server.enabled %}<span class="badge badge--danger">disabled</span>{% endif %}
+      {% if not server.public %}<span class="badge">restricted</span>{% endif %}
+      {% if server.allow_private %}<span class="badge">private network</span>{% endif %}
+      {% if server.protocol_version %}
+      <span class="badge badge--leaf">MCP {{ server.protocol_version }}</span>
+      {% endif %}
+    </div>
+    <code class="model-row__id">{{ server.slug }} · {{ server.url }}</code>
+    {% if message %}
+    <p class="text-xs {{ 'danger' if message_kind == 'error' else 'faint' }}">{{ message }}</p>
+    {% elif server.last_error %}
+    <p class="text-xs danger">{{ server.last_error }}</p>
+    {% endif %}
+  </div>
+
+  <div class="model-row__actions">
+    <button class="btn btn--sm" type="button"
+            hx-post="/admin/mcp/{{ server.id }}/test"
+            hx-target="#mcp-{{ server.id }}" hx-swap="outerHTML">
+      {{ icon("refresh", "icon--sm") }} Test &amp; refresh
+    </button>
+    <a class="btn btn--sm" href="/admin/mcp/{{ server.id }}/edit">Edit</a>
+  </div>
+</div>

src/lembas/web/templates/admin/mcp.html

@@ -0,0 +1,53 @@
+{% extends "admin/_layout.html" %}
+{% from "_macros.html" import icon %}
+{% set section = "mcp" %}
+
+{% block title %}MCP servers - LLeMbas{% endblock %}
+{% block heading %}MCP servers{% endblock %}
+
+{% block admin_content %}
+<p class="admin-lede">
+  Remote servers speaking the Model Context Protocol over HTTP. Their tools are
+  offered beside the built-in ones to models marked <strong>MCP servers</strong>.
+  The list of tools is discovered and cached — press <strong>Test &amp;
+  refresh</strong> after adding one, and again whenever the server changes.
+</p>
+
+<div class="alert">
+  {{ icon("shield", "icon--sm") }}
+  <span>
+    A server's tool names and descriptions are sent to the model as
+    instructions, and what it returns is read back as fact. Add servers you
+    trust, the way you would a dependency.
+  </span>
+</div>
+
+{% if saved %}
+<div class="alert alert--success">{{ icon("check", "icon--sm") }} <span>{{ saved }}</span></div>
+{% endif %}
+
+<div class="btn-row">
+  <a class="btn btn--primary" href="/admin/mcp/new">
+    {{ icon("plus", "icon--sm") }} Add a server
+  </a>
+</div>
+
+{% if not servers %}
+<div class="empty" style="padding: var(--sp-10) 0">
+  {{ icon("server", "empty__mark") }}
+  <p class="empty__text">
+    No servers yet. You will need the URL of an MCP endpoint that speaks
+    streamable HTTP — local ones launched as a subprocess are not supported.
+  </p>
+</div>
+{% else %}
+
+<div class="model-rows">
+  {% for server in servers %}
+  {% with tool_count = counts[server.id] %}
+  {% include "admin/_mcp_row.html" %}
+  {% endwith %}
+  {% endfor %}
+</div>
+{% endif %}
+{% endblock %}

src/lembas/web/templates/admin/mcp_detail.html

@@ -0,0 +1,241 @@
+{% extends "admin/_layout.html" %}
+{% from "_macros.html" import icon %}
+{% set section = "mcp" %}
+
+{% block title %}{{ "New server" if is_new else server.name }} - LLeMbas{% endblock %}
+{% block heading %}{{ "New MCP server" if is_new else server.name }}{% endblock %}
+
+{% block admin_content %}
+<nav class="crumbs">
+  <a class="crumbs__back" href="/admin/mcp">
+    {{ icon("chevron-right", "icon--sm crumbs__icon") }} All servers
+  </a>
+</nav>
+
+{% if error %}
+<div class="alert alert--error">{{ icon("warning", "icon--sm") }} <span>{{ error }}</span></div>
+{% endif %}
+
+{% if server.last_error %}
+<div class="alert alert--error">
+  {{ icon("warning", "icon--sm") }}
+  <span>Last contacted unsuccessfully: {{ server.last_error }}</span>
+</div>
+{% endif %}
+
+<form method="post" action="{{ '/admin/mcp' if is_new else '/admin/mcp/' ~ server.id }}"
+      class="form-grid">
+
+  <section class="card">
+    <h2 class="card__title">The server</h2>
+
+    <div class="field">
+      <label class="field__label" for="name">Name</label>
+      <input class="input" id="name" name="name" value="{{ server.name }}" required
+             maxlength="120" placeholder="GitHub">
+    </div>
+
+    <div class="field">
+      <label class="field__label" for="slug">Identifier</label>
+      <input class="input input--mono" id="slug" name="slug" value="{{ server.slug }}" required
+             maxlength="24" pattern="[a-z0-9][a-z0-9_\-]*" placeholder="github">
+      <p class="field__hint">
+        Prefixed onto every tool name this server offers, so that two servers
+        both exposing <code>search</code> do not collide.
+      </p>
+    </div>
+
+    <div class="field">
+      <label class="field__label" for="url">Endpoint URL</label>
+      <input class="input input--mono" id="url" name="url" value="{{ server.url }}" required
+             placeholder="https://mcp.example.com/mcp">
+      <p class="field__hint">
+        The streamable-HTTP endpoint itself, the one that accepts a POST. A
+        server that answers with a redirect to somewhere else will be refused.
+      </p>
+    </div>
+
+    <div class="field">
+      <label class="field__label" for="headers">Extra headers</label>
+      <textarea class="textarea input--mono" id="headers" name="headers" rows="3"
+                spellcheck="false">{{ headers_text }}</textarea>
+      <p class="field__hint">One <code>Name: value</code> per line.</p>
+    </div>
+
+    <div class="field">
+      <label class="field__label" for="timeout">Timeout (seconds)</label>
+      <input class="input" id="timeout" name="timeout" value="{{ server.timeout }}"
+             inputmode="numeric">
+    </div>
+
+    <div class="field">
+      <label class="field__label" for="max_chars">Most characters to keep per call</label>
+      <input class="input" id="max_chars" name="max_chars" value="{{ server.max_chars }}"
+             inputmode="numeric">
+    </div>
+  </section>
+
+  <section class="card">
+    <h2 class="card__title">Credential</h2>
+
+    <div class="field">
+      <label class="field__label" for="secret_placement">How it is sent</label>
+      <select class="select" id="secret_placement" name="secret_placement">
+        {% for value, label in secret_placements %}
+        <option value="{{ value }}" {{ 'selected' if value == server.secret_placement }}>
+          {{ label }}
+        </option>
+        {% endfor %}
+      </select>
+    </div>
+
+    <div class="field">
+      <label class="field__label" for="secret_name">Header or parameter name</label>
+      <input class="input input--mono" id="secret_name" name="secret_name"
+             value="{{ server.secret_name }}" maxlength="120">
+    </div>
+
+    <div class="field">
+      <label class="field__label" for="secret">Secret</label>
+      <input class="input input--mono" id="secret" name="secret" type="password"
+             autocomplete="off" placeholder="No secret set"
+             value="{{ unchanged if server.secret_encrypted else '' }}">
+      <p class="field__hint">
+        {% if server.secret_encrypted %}
+          Currently <code>{{ masked }}</code>. Leave the dots alone to keep it,
+          or clear the field to remove it.
+        {% else %}
+          Encrypted at rest and never shown again.
+        {% endif %}
+      </p>
+    </div>
+  </section>
+
+  {% if tools %}
+  <section class="card">
+    <h2 class="card__title">Tools it offers</h2>
+    <p class="field__hint">
+      Discovered at the last refresh. Untick one to withhold it — a tool this
+      server adds later is offered by default.
+    </p>
+
+    <input type="hidden" name="tool_choices" value="1">
+    <div class="checkbox-row checkbox-row--stacked">
+      {% for tool in tools %}
+      <label class="checkbox">
+        <input type="hidden" name="tool_names" value="{{ tool.name }}">
+        <input type="checkbox" name="tool_names_on" value="{{ tool.name }}"
+               {{ 'checked' if tool.on }}>
+        <span>
+          <code>{{ tool.offer_name or tool.name }}</code>
+          {% if tool.offer_name and tool.offer_name != tool.name %}
+          <span class="faint text-xs">({{ tool.name }} on the server)</span>
+          {% endif %}
+          {% if tool.description %}
+          <br><span class="faint text-xs">{{ tool.description }}</span>
+          {% endif %}
+        </span>
+      </label>
+      {% endfor %}
+    </div>
+  </section>
+  {% elif not is_new %}
+  <section class="card">
+    <h2 class="card__title">Tools it offers</h2>
+    <p class="field__hint">
+      Nothing discovered yet. Save, then press <strong>Test &amp; refresh</strong>
+      on the <a href="/admin/mcp">list</a>.
+    </p>
+  </section>
+  {% endif %}
+
+  <section class="card">
+    <h2 class="card__title">Guidance</h2>
+
+    <div class="field">
+      <textarea class="textarea" name="guidance" rows="4"
+                placeholder="- Use the GitHub tools for anything about our repositories."
+                >{{ server.guidance }}</textarea>
+      <p class="field__hint">
+        Added to the system message whenever any tool from this server is
+        offered. One piece of guidance for the server, not one per tool —
+        the tools carry their own descriptions.
+        {% if prompt_overridden %}
+          <br><strong>Someone has overridden this wording under
+          <a href="/admin/prompts">Prompts</a></strong> — that is what the model
+          sees, not this.
+        {% endif %}
+      </p>
+    </div>
+  </section>
+
+  <section class="card">
+    <h2 class="card__title">Availability</h2>
+
+    <div class="field">
+      <div class="checkbox-row">
+        <label class="checkbox">
+          <input type="checkbox" name="enabled" value="true" {{ 'checked' if server.enabled }}>
+          <span>Enabled — offered in chats</span>
+        </label>
+        <label class="checkbox">
+          <input type="checkbox" name="allow_private" value="true"
+                 {{ 'checked' if server.allow_private }}>
+          <span>May reach private and loopback addresses</span>
+        </label>
+      </div>
+      <p class="field__hint">
+        Tick the second only for a server on your own network. It is what stops
+        this being aimed at LLeMbas itself, a router, or a metadata endpoint.
+      </p>
+    </div>
+
+    <div class="field">
+      <label class="checkbox">
+        <input type="checkbox" name="public" value="true" {{ 'checked' if server.public }}>
+        <span>Available to everyone</span>
+      </label>
+    </div>
+
+    <div class="field">
+      <span class="field__label">Groups with access</span>
+      {% if groups %}
+      <div class="checkbox-row">
+        {% for group in groups %}
+        <label class="checkbox">
+          <input type="checkbox" name="group_ids" value="{{ group.id }}"
+                 {{ 'checked' if group.id in selected_groups }}>
+          <span>{{ group.name }}</span>
+        </label>
+        {% endfor %}
+      </div>
+      <p class="field__hint">Ignored while the server is available to everyone.</p>
+      {% else %}
+      <p class="field__hint">
+        No groups yet — <a href="/admin/groups">create one</a> to restrict access.
+      </p>
+      {% endif %}
+    </div>
+
+    <div class="field">
+      <label class="field__label" for="position">Position</label>
+      <input class="input" id="position" name="position" value="{{ server.position }}"
+             inputmode="numeric">
+    </div>
+  </section>
+
+  <div class="btn-row">
+    <button class="btn btn--primary" type="submit">
+      {{ "Add server" if is_new else "Save changes" }}
+    </button>
+    <a class="btn btn--ghost" href="/admin/mcp">Back to all servers</a>
+    {% if not is_new %}
+    <button class="btn btn--danger" type="submit" formnovalidate
+            formaction="/admin/mcp/{{ server.id }}/delete"
+            data-confirm-button="Delete the server “{{ server.name }}”? Chats that used its tools keep their transcripts.">
+      Delete
+    </button>
+    {% endif %}
+  </div>
+</form>
+{% endblock %}

tests/test_admin_tools.py

@@ -301,3 +301,196 @@ def test_the_secret_survives_a_round_trip_through_the_form(client: TestClient, d
     )
     db.refresh(tool)
     assert decrypt(tool.secret_encrypted) == "s3cret"
+
+
+# --- MCP servers -------------------------------------------------------------
+def _mcp_form(**overrides) -> dict:
+    base = {
+        "name": "GitHub",
+        "slug": "github",
+        "url": "https://mcp.test/rpc",
+        "headers": "",
+        "secret_placement": "none",
+        "secret_name": "Authorization",
+        "timeout": "30",
+        "max_chars": "8000",
+        "position": "0",
+        "enabled": "true",
+        "public": "true",
+    }
+    base.update(overrides)
+    return {key: value for key, value in base.items() if value is not None}
+
+
+@pytest.fixture
+def fake_mcp(mock_http, monkeypatch):
+    """A server that answers initialize and tools/list.
+
+    Invented hostnames resolve to a public address; a literal IP is handed back
+    as itself, so a test about a private address is still testing one.
+    """
+    import ipaddress
+
+    def resolve(host, *_args, **_kwargs):
+        try:
+            ipaddress.ip_address(host)
+        except ValueError:
+            return [(2, 1, 6, "", ("93.184.216.34", 80))]
+        return [(2, 1, 6, "", (host, 80))]
+
+    monkeypatch.setattr("socket.getaddrinfo", resolve)
+
+    def handler(request: httpx.Request) -> httpx.Response:
+        if request.method != "POST":
+            return httpx.Response(405)
+        message = json.loads(request.content)
+        if message.get("method") == "notifications/initialized":
+            return httpx.Response(202)
+        if message.get("method") == "initialize":
+            return httpx.Response(
+                200,
+                json={
+                    "jsonrpc": "2.0",
+                    "id": message["id"],
+                    "result": {"protocolVersion": "2025-06-18", "serverInfo": {"name": "fake"}},
+                },
+                headers={"mcp-session-id": "s-1"},
+            )
+        return httpx.Response(
+            200,
+            json={
+                "jsonrpc": "2.0",
+                "id": message["id"],
+                "result": {"tools": [{"name": "search", "description": "Search."}]},
+            },
+        )
+
+    mock_http(handler)
+    return handler
+
+
+def test_the_mcp_pages_are_refused_to_a_plain_user(client: TestClient, plain_user):
+    assert client.get("/admin/mcp").status_code == 403
+    assert client.get("/admin/mcp/new").status_code == 403
+    assert client.post("/admin/mcp", data=_mcp_form()).status_code == 403
+
+
+def test_mcp_new_is_not_parsed_as_a_server_id(client: TestClient, registered):
+    response = client.get("/admin/mcp/new")
+    assert response.status_code == 200
+    assert "New MCP server" in response.text
+
+
+def test_adding_a_server_discovers_its_tools(client: TestClient, db, registered, fake_mcp):
+    from lembas.db.models import McpServer
+
+    client.post("/admin/mcp", data=_mcp_form(), follow_redirects=False)
+
+    server = db.scalar(select(McpServer))
+    assert server.slug == "github"
+    assert [entry["name"] for entry in server.tools_json] == ["search"]
+    assert server.tools_json[0]["offer_name"] == "github_search"
+    assert server.protocol_version == "2025-06-18"
+
+
+def test_refreshing_a_server_swaps_its_row(client: TestClient, db, registered, fake_mcp):
+    from lembas.db.models import McpServer
+
+    client.post("/admin/mcp", data=_mcp_form(), follow_redirects=False)
+    server = db.scalar(select(McpServer))
+
+    response = client.post(f"/admin/mcp/{server.id}/test")
+    assert response.status_code == 200
+    assert "found 1 tool." in response.text
+    assert f'id="mcp-{server.id}"' in response.text, "the fragment must render standalone"
+
+
+def test_a_server_that_cannot_be_reached_says_so(
+    client: TestClient, db, registered, mock_http, monkeypatch
+):
+    from lembas.db.models import McpServer
+
+    monkeypatch.setattr(
+        "socket.getaddrinfo", lambda *a, **k: [(2, 1, 6, "", ("93.184.216.34", 80))]
+    )
+    mock_http(lambda _r: httpx.Response(500, text="down"))
+
+    response = client.post("/admin/mcp", data=_mcp_form(), follow_redirects=False)
+    assert response.status_code == 303
+    server = db.scalar(select(McpServer))
+    assert server is not None, "the row is still saved so the URL can be corrected"
+    assert server.last_error
+
+
+def test_a_private_url_is_not_contacted_unless_the_box_is_ticked(
+    client: TestClient, db, registered, fake_mcp
+):
+    """The row still saves, so the URL can be corrected -- but the discovery
+    that runs straight after it is refused, and the row says why."""
+    from lembas.db.models import McpServer
+
+    client.post(
+        "/admin/mcp", data=_mcp_form(url="http://127.0.0.1:9000/rpc"), follow_redirects=False
+    )
+    server = db.scalar(select(McpServer))
+    assert server.allow_private is False
+    assert "private or local" in server.last_error
+    assert server.tools_json == []
+
+    client.post(
+        f"/admin/mcp/{server.id}",
+        data=_mcp_form(url="http://127.0.0.1:9000/rpc", allow_private="true"),
+        follow_redirects=False,
+    )
+    client.post(f"/admin/mcp/{server.id}/test")
+    db.refresh(server)
+    assert server.last_error == ""
+    assert [entry["name"] for entry in server.tools_json] == ["search"]
+
+
+def test_a_duplicate_mcp_slug_is_refused(client: TestClient, db, registered, fake_mcp):
+    from lembas.db.models import McpServer
+
+    client.post("/admin/mcp", data=_mcp_form(), follow_redirects=False)
+    response = client.post("/admin/mcp", data=_mcp_form(name="Other"), follow_redirects=False)
+    assert "already a server" in response.text
+    assert len(list(db.scalars(select(McpServer)))) == 1
+
+
+def test_a_url_that_is_not_http_is_refused(client: TestClient, db, registered):
+    from lembas.db.models import McpServer
+
+    response = client.post(
+        "/admin/mcp", data=_mcp_form(url="ftp://mcp.test/rpc"), follow_redirects=False
+    )
+    assert "http" in response.text
+    assert db.scalar(select(McpServer)) is None
+
+
+def test_unticking_a_tool_withholds_it(client: TestClient, db, registered, fake_mcp):
+    from lembas.db.models import McpServer, User
+    from lembas.services.mcp import registry as mcp_registry
+
+    client.post("/admin/mcp", data=_mcp_form(), follow_redirects=False)
+    server = db.scalar(select(McpServer))
+
+    client.post(
+        f"/admin/mcp/{server.id}",
+        data={**_mcp_form(), "tool_choices": "1", "tool_names": "search"},
+        follow_redirects=False,
+    )
+    db.refresh(server)
+    assert server.tool_overrides_json == {"search": False}
+
+    owner = db.scalar(select(User))
+    assert mcp_registry.tool_defs(db, owner) == []
+
+
+def test_deleting_a_server_removes_it(client: TestClient, db, registered, fake_mcp):
+    from lembas.db.models import McpServer
+
+    client.post("/admin/mcp", data=_mcp_form(), follow_redirects=False)
+    server = db.scalar(select(McpServer))
+
+    client.post(f"/admin/mcp/{server.id}/delete", follow_redirects=False)
+    assert db.scalar(select(McpServer)) is None

tests/test_generation_tools.py

@@ -207,3 +207,126 @@ async def _empty_search(_config, _query, *, limit=None):
 async def _never_called_title(*_args, **_kwargs):
     """Auto-titling makes its own request; these tests are about the tool loop."""
     return "A title"
+
+
+# --- Progress and concurrency ------------------------------------------------
+async def test_the_status_names_the_running_tool_and_is_cleared(db, user_id, monkeypatch):
+    """A remote tool can take seconds with nothing streaming, and a silent
+    pause is exactly what a hang looks like."""
+    settings_store.update(db, {"enabled": True}, key=settings_store.SEARCH)
+    chat_id, message_id = _chat_with_tools(db, user_id)
+
+    seen: list[str] = []
+
+    async def fake_search(_config, _query, *, limit=None):
+        seen.append(generation.status)
+        return []
+
+    monkeypatch.setattr("lembas.services.search.run", fake_search)
+    monkeypatch.setattr(
+        generation_service,
+        "stream_chat",
+        _stub_stream(
+            [[_tool_call_chunk("web_search", '{"query": "mallorn"}')], [_text_chunk("Done.")]],
+            [],
+        ),
+    )
+
+    generation = generation_service.Generation(chat_id=chat_id, message_id=message_id)
+    await generation_service._run(generation)
+
+    assert seen == ["Running web_search…"]
+    assert generation.status == "", "and it is cleared once they are done"
+
+
+async def test_results_stay_paired_with_their_calls_when_run_together(db, user_id, monkeypatch):
+    """Indexed rather than appended as they finish: an endpoint matching on
+    tool_call_id would otherwise pair the right id with the wrong content."""
+    import asyncio
+
+    settings_store.update(db, {"enabled": True}, key=settings_store.SEARCH)
+    chat_id, message_id = _chat_with_tools(db, user_id)
+
+    async def slow_first(_config, query, *, limit=None):
+        # The first call finishes last, which is the whole point of the test.
+        await asyncio.sleep(0.02 if query == "first" else 0)
+        return [SearchResult(f"result for {query}", f"https://t.test/{query}", "")]
+
+    monkeypatch.setattr("lembas.services.search.run", slow_first)
+
+    two_calls = {
+        "choices": [
+            {"delta": {"tool_calls": [
+                {"index": 0, "id": "a", "function": {
+                    "name": "web_search", "arguments": '{"query": "first"}'}},
+                {"index": 1, "id": "b", "function": {
+                    "name": "web_search", "arguments": '{"query": "second"}'}},
+            ]}}
+        ]
+    }
+    payloads: list[dict] = []
+    monkeypatch.setattr(
+        generation_service,
+        "stream_chat",
+        _stub_stream([[two_calls], [_text_chunk("Done.")]], payloads),
+    )
+
+    generation = generation_service.Generation(chat_id=chat_id, message_id=message_id)
+    await generation_service._run(generation)
+
+    turns = [m for m in payloads[1]["messages"] if m.get("role") == "tool"]
+    assert [turn["tool_call_id"] for turn in turns] == ["a", "b"]
+    assert "first" in turns[0]["content"] and "second" in turns[1]["content"]
+    # And the transcript keeps the same order.
+    assert [event["query"] for event in generation.tool_events] == ["first", "second"]
+
+
+async def test_a_custom_tool_runs_inside_the_loop(db, user_id, monkeypatch, mock_http):
+    """End to end: a row becomes an offered schema, the model calls it, and the
+    result comes back in the next request's messages."""
+    import httpx
+
+    from lembas.db.models import CustomTool
+
+    monkeypatch.setattr(
+        "socket.getaddrinfo", lambda *a, **k: [(2, 1, 6, "", ("93.184.216.34", 80))]
+    )
+    mock_http(lambda _r: httpx.Response(200, json={"summary": "Sunny in Minas Tirith."}))
+
+    db.add(
+        CustomTool(
+            slug="weather",
+            name="Weather",
+            description="Look up the weather.",
+            url_template="https://api.test/{{city}}",
+            parameters_json={"type": "object", "properties": {"city": {"type": "string"}}},
+            response_mode="json",
+            response_path="summary",
+        )
+    )
+    db.commit()
+
+    chat_id, message_id = _chat_with_tools(db, user_id)
+    payloads: list[dict] = []
+    monkeypatch.setattr(
+        generation_service,
+        "stream_chat",
+        _stub_stream(
+            [
+                [_tool_call_chunk("weather", '{"city": "Minas Tirith"}')],
+                [_text_chunk("It is sunny.")],
+            ],
+            payloads,
+        ),
+    )
+
+    generation = generation_service.Generation(chat_id=chat_id, message_id=message_id)
+    await generation_service._run(generation)
+
+    offered = {tool["function"]["name"] for tool in payloads[0]["tools"]}
+    assert "weather" in offered
+
+    tool_turns = [m for m in payloads[1]["messages"] if m.get("role") == "tool"]
+    assert tool_turns[0]["content"] == "Sunny in Minas Tirith."
+    assert generation.tool_events[0]["kind"] == "custom"
+    assert generation.tool_events[0]["label"] == "Weather"

tests/test_mcp.py

@@ -0,0 +1,506 @@
+"""The MCP client: framing, the session lifecycle, and what comes back.
+
+The fake server below answers the way a real one does -- JSON-RPC over POST,
+either as one JSON object or as an event stream, with a session id it expects
+echoed. Everything is driven through `mock_http`, so no test touches a network.
+"""
+
+from __future__ import annotations
+
+import json
+
+import httpx
+import pytest
+from sqlalchemy import select
+
+from lembas.db.models import McpServer
+from lembas.services.mcp import client, protocol, registry
+from lembas.services.mcp.protocol import McpError
+
+
+@pytest.fixture(autouse=True)
+def dns(monkeypatch):
+    """`mcp.test` does not exist and check_url resolves for real."""
+    import ipaddress
+    import socket
+
+    def resolve(host, *_args, **_kwargs):
+        try:
+            ipaddress.ip_address(host)
+        except ValueError:
+            return [(2, 1, 6, "", ("93.184.216.34", 80))]
+        return [(2, 1, 6, "", (host, 80))]
+
+    monkeypatch.setattr(socket, "getaddrinfo", resolve)
+
+
+def _spec(**overrides) -> client.McpSpec:
+    return client.McpSpec(
+        **{"slug": "github", "name": "GitHub", "url": "https://mcp.test/rpc", **overrides}
+    )
+
+
+class FakeServer:
+    """One server's worth of behaviour, as an httpx handler."""
+
+    def __init__(self, *, tools=None, stream=False, session_id="s-1", pages=None):
+        self.tools = tools if tools is not None else [
+            {"name": "search", "description": "Search things.", "inputSchema": {"type": "object"}}
+        ]
+        self.stream = stream
+        self.session_id = session_id
+        self.pages = pages
+        self.requests: list[dict] = []
+        self.headers: list[httpx.Headers] = []
+        self.result_for = {}
+
+    def _frame(self, message: dict) -> httpx.Response:
+        if self.stream:
+            body = "".join(f"data: {line}\n" for line in json.dumps(message).splitlines()) + "\n"
+            return httpx.Response(
+                200, text=body, headers={"content-type": "text/event-stream"}
+            )
+        return httpx.Response(200, json=message)
+
+    def __call__(self, request: httpx.Request) -> httpx.Response:
+        self.headers.append(request.headers)
+        if request.method == "DELETE":
+            return httpx.Response(405)
+
+        message = json.loads(request.content)
+        self.requests.append(message)
+        method, request_id = message.get("method"), message.get("id")
+
+        if method == "notifications/initialized":
+            return httpx.Response(202)
+
+        if method == "initialize":
+            headers = {"mcp-session-id": self.session_id} if self.session_id else {}
+            response = self._frame(
+                {
+                    "jsonrpc": "2.0",
+                    "id": request_id,
+                    "result": {
+                        "protocolVersion": "2025-06-18",
+                        "capabilities": {"tools": {}},
+                        "serverInfo": {"name": "fake", "version": "1"},
+                    },
+                }
+            )
+            return httpx.Response(
+                200, content=response.content, headers={**dict(response.headers), **headers}
+            )
+
+        if method == "tools/list":
+            if self.pages is not None:
+                cursor = (message.get("params") or {}).get("cursor", "")
+                index = int(cursor or 0)
+                page = self.pages[index]
+                result = {"tools": page}
+                if index + 1 < len(self.pages):
+                    result["nextCursor"] = str(index + 1)
+                return self._frame({"jsonrpc": "2.0", "id": request_id, "result": result})
+            return self._frame(
+                {"jsonrpc": "2.0", "id": request_id, "result": {"tools": self.tools}}
+            )
+
+        if method == "tools/call":
+            name = (message.get("params") or {}).get("name")
+            result = self.result_for.get(
+                name, {"content": [{"type": "text", "text": "It worked."}]}
+            )
+            return self._frame({"jsonrpc": "2.0", "id": request_id, "result": result})
+
+        return self._frame(
+            {"jsonrpc": "2.0", "id": request_id, "error": {"code": -32601, "message": "Unknown"}}
+        )
+
+
+# --- The lifecycle -----------------------------------------------------------
+async def test_initialize_is_followed_by_the_initialized_notification(mock_http):
+    server = FakeServer()
+    mock_http(server)
+
+    tools, info, version = await client.list_tools(_spec())
+    methods = [request.get("method") for request in server.requests]
+    assert methods[:3] == ["initialize", "notifications/initialized", "tools/list"]
+    assert info == {"name": "fake", "version": "1"}
+    assert version == "2025-06-18"
+    assert [tool["name"] for tool in tools] == ["search"]
+
+
+async def test_the_session_id_is_echoed_on_every_later_request(mock_http):
+    server = FakeServer(session_id="abc-123")
+    mock_http(server)
+
+    await client.list_tools(_spec())
+    assert "mcp-session-id" not in server.headers[0]
+    assert all(headers["mcp-session-id"] == "abc-123" for headers in server.headers[1:])
+    assert all("mcp-protocol-version" in headers for headers in server.headers[1:])
+
+
+async def test_a_server_that_issues_no_session_is_fine(mock_http):
+    mock_http(FakeServer(session_id=""))
+    tools, _info, _version = await client.list_tools(_spec())
+    assert tools
+
+
+async def test_an_expired_session_is_reinitialised_once(mock_http):
+    """A 404 on a request carrying a session id means the server forgot us."""
+    server = FakeServer()
+    calls = {"list": 0}
+    inner = server.__call__
+
+    def handler(request: httpx.Request) -> httpx.Response:
+        if request.method == "POST":
+            message = json.loads(request.content)
+            if message.get("method") == "tools/list":
+                calls["list"] += 1
+                if calls["list"] == 1:
+                    return httpx.Response(404, text="no such session")
+        return inner(request)
+
+    mock_http(handler)
+    tools, _info, _version = await client.list_tools(_spec())
+    assert tools, "the retry after re-initialising must succeed"
+    assert [r.get("method") for r in server.requests].count("initialize") == 2
+
+
+async def test_a_json_response_and_an_event_stream_parse_the_same(mock_http):
+    plain = FakeServer(stream=False)
+    mock_http(plain)
+    from_json, _i, _v = await client.list_tools(_spec())
+
+    streamed = FakeServer(stream=True)
+    mock_http(streamed)
+    from_stream, _i, _v = await client.list_tools(_spec())
+
+    assert from_json == from_stream
+
+
+async def test_tools_list_follows_the_cursor(mock_http):
+    server = FakeServer(
+        pages=[
+            [{"name": "one", "inputSchema": {"type": "object"}}],
+            [{"name": "two", "inputSchema": {"type": "object"}}],
+        ]
+    )
+    mock_http(server)
+    tools, _i, _v = await client.list_tools(_spec())
+    assert [tool["name"] for tool in tools] == ["one", "two"]
+
+
+async def test_a_jsonrpc_error_becomes_a_readable_message(mock_http):
+    def handler(request: httpx.Request) -> httpx.Response:
+        message = json.loads(request.content)
+        if message.get("method") == "notifications/initialized":
+            return httpx.Response(202)
+        return httpx.Response(
+            200,
+            json={
+                "jsonrpc": "2.0",
+                "id": message.get("id"),
+                "error": {"code": -32602, "message": "invalid params"},
+            },
+        )
+
+    mock_http(handler)
+    with pytest.raises(McpError, match="invalid params"):
+        await client.list_tools(_spec())
+
+
+async def test_an_http_error_is_reported_with_its_status(mock_http):
+    mock_http(lambda _r: httpx.Response(502, text="bad gateway"))
+    with pytest.raises(McpError, match="502"):
+        await client.list_tools(_spec())
+
+
+# --- Reaching the network ----------------------------------------------------
+async def test_the_url_is_checked_before_the_first_request(mock_http):
+    server = FakeServer()
+    mock_http(server)
+    from lembas.services.fetch import FetchError
+
+    with pytest.raises(FetchError, match="private or local"):
+        await client.list_tools(_spec(url="http://127.0.0.1:9000/rpc"))
+    assert server.requests == [], "nothing may be sent before the address is checked"
+
+
+async def test_a_302_is_refused_and_a_307_is_followed_and_checked(mock_http):
+    """301, 302 and 303 turn a POST into a GET, which means nothing to a
+    JSON-RPC endpoint."""
+    inner = FakeServer()
+
+    def redirect_302(request: httpx.Request) -> httpx.Response:
+        return httpx.Response(302, headers={"location": "https://elsewhere.test/rpc"})
+
+    mock_http(redirect_302)
+    with pytest.raises(McpError, match="turn the request into a GET"):
+        await client.list_tools(_spec())
+
+    def redirect_307(request: httpx.Request) -> httpx.Response:
+        if request.url.host == "mcp.test":
+            return httpx.Response(307, headers={"location": "https://elsewhere.test/rpc"})
+        return inner(request)
+
+    mock_http(redirect_307)
+    tools, _i, _v = await client.list_tools(_spec())
+    assert tools
+
+
+async def test_a_redirect_to_a_private_address_is_refused(mock_http, monkeypatch):
+    def resolve(host, *_args, **_kwargs):
+        if host == "inside.test":
+            return [(2, 1, 6, "", ("10.0.0.5", 80))]
+        return [(2, 1, 6, "", ("93.184.216.34", 80))]
+
+    monkeypatch.setattr("socket.getaddrinfo", resolve)
+    mock_http(lambda _r: httpx.Response(307, headers={"location": "https://inside.test/rpc"}))
+
+    from lembas.services.fetch import FetchError
+
+    with pytest.raises(FetchError, match="private or local"):
+        await client.list_tools(_spec())
+
+
+async def test_the_secret_reaches_the_server(mock_http):
+    server = FakeServer()
+    mock_http(server)
+    await client.list_tools(_spec(secret="tok", secret_placement="bearer"))
+    assert server.headers[0]["authorization"] == "Bearer tok"
+
+
+# --- Reading a tool result ---------------------------------------------------
+async def test_text_blocks_are_joined(mock_http):
+    server = FakeServer()
+    server.result_for["search"] = {
+        "content": [{"type": "text", "text": "one"}, {"type": "text", "text": "two"}]
+    }
+    mock_http(server)
+    text, failed = await client.call_tool(_spec(), "search", {})
+    assert text == "one\n\ntwo"
+    assert failed is False
+
+
+async def test_an_image_block_is_described_rather_than_forwarded(mock_http):
+    """A tool turn is a string, images only reach models marked as having
+    vision, and base64 in a tool result fills a window with nothing."""
+    server = FakeServer()
+    server.result_for["search"] = {
+        "content": [{"type": "image", "mimeType": "image/png", "data": "A" * 400}]
+    }
+    mock_http(server)
+    text, _failed = await client.call_tool(_spec(), "search", {})
+    assert "A" * 40 not in text
+    assert "image/png" in text
+    assert "not shown to the model" in text
+
+
+async def test_a_resource_block_with_text_keeps_it(mock_http):
+    server = FakeServer()
+    server.result_for["search"] = {
+        "content": [
+            {"type": "resource", "resource": {"uri": "file:///a.txt", "text": "contents"}}
+        ]
+    }
+    mock_http(server)
+    text, _failed = await client.call_tool(_spec(), "search", {})
+    assert "file:///a.txt" in text and "contents" in text
+
+
+async def test_a_resource_block_with_no_text_names_its_type(mock_http):
+    server = FakeServer()
+    server.result_for["search"] = {
+        "content": [
+            {"type": "resource", "resource": {"uri": "file:///a.bin", "mimeType": "application/x"}}
+        ]
+    }
+    mock_http(server)
+    text, _failed = await client.call_tool(_spec(), "search", {})
+    assert "application/x" in text
+
+
+async def test_an_unknown_block_type_degrades_to_a_note(mock_http):
+    server = FakeServer()
+    server.result_for["search"] = {"content": [{"type": "hologram"}]}
+    mock_http(server)
+    text, _failed = await client.call_tool(_spec(), "search", {})
+    assert "hologram" in text
+
+
+async def test_structured_content_is_used_when_there_is_no_text(mock_http):
+    server = FakeServer()
+    server.result_for["search"] = {"content": [], "structuredContent": {"count": 3}}
+    mock_http(server)
+    text, _failed = await client.call_tool(_spec(), "search", {})
+    assert json.loads(text) == {"count": 3}
+
+
+async def test_is_error_is_reported_as_a_failure(mock_http):
+    server = FakeServer()
+    server.result_for["search"] = {
+        "content": [{"type": "text", "text": "no such repo"}],
+        "isError": True,
+    }
+    mock_http(server)
+    _text, failed = await client.call_tool(_spec(), "search", {})
+    assert failed is True
+
+
+# --- Bounding what a server can say ------------------------------------------
+def test_a_giant_description_is_truncated():
+    cleaned = protocol.clean_tool({"name": "x", "description": "d" * 9000})
+    assert len(cleaned["description"]) == protocol.MAX_DESCRIPTION
+
+
+def test_a_schema_that_is_not_an_object_is_replaced():
+    cleaned = protocol.clean_tool({"name": "x", "inputSchema": "nope"})
+    assert cleaned["schema"] == {"type": "object", "properties": {}}
+
+
+def test_a_giant_schema_is_replaced():
+    huge = {"type": "object", "properties": {f"p{i}": {"type": "string"} for i in range(1000)}}
+    cleaned = protocol.clean_tool({"name": "x", "inputSchema": huge})
+    assert cleaned["schema"] == {"type": "object", "properties": {}}
+
+
+def test_an_entry_with_no_name_is_dropped():
+    assert protocol.clean_tool({"description": "orphan"}) is None
+    assert protocol.clean_tool("not a dict") is None
+
+
+# --- Namespacing -------------------------------------------------------------
+def test_tool_names_are_namespaced_and_within_the_charset():
+    taken: set[str] = set()
+    assert registry.offer_name("github", "search", taken=taken) == "github_search"
+    assert registry.offer_name("github", "Create Issue!", taken=taken) == "github_create_issue"
+    assert all(registry.FUNCTION_NAME.match(name) for name in taken)
+
+
+def test_a_name_colliding_with_a_builtin_is_renamed():
+    from lembas.services import tools as tools_service
+
+    taken = set(tools_service.REGISTRY)
+    name = registry.offer_name("", "web_search", taken=taken)
+    assert name != "web_search"
+    assert name not in tools_service.REGISTRY
+
+
+def test_a_truncation_collision_is_disambiguated():
+    """Two long names can collide once cut to 64 characters where the full ones
+    would not have."""
+    taken: set[str] = set()
+    first = registry.offer_name("srv", "a" * 200, taken=taken)
+    second = registry.offer_name("srv", "a" * 201, taken=taken)
+    assert first != second
+    assert len(first) <= 64 and len(second) <= 64
+
+
+# --- Rows become tools -------------------------------------------------------
+def _server(db, **overrides) -> McpServer:
+    row = McpServer(
+        **{
+            "slug": "github",
+            "name": "GitHub",
+            "url": "https://mcp.test/rpc",
+            "tools_json": [
+                {
+                    "name": "search",
+                    "offer_name": "github_search",
+                    "description": "Search things.",
+                    "schema": {"type": "object", "properties": {}},
+                }
+            ],
+            **overrides,
+        }
+    )
+    db.add(row)
+    db.commit()
+    return row
+
+
+def test_a_cached_tool_becomes_a_definition(db, user_id):
+    from lembas.db.models import User
+
+    _server(db)
+    defs = registry.tool_defs(db, db.get(User, user_id))
+    assert [tool.name for tool in defs] == ["github_search"]
+    assert defs[0].family == "mcp:github"
+
+
+def test_a_tool_switched_off_is_not_offered(db, user_id):
+    from lembas.db.models import User
+
+    _server(db, tool_overrides_json={"search": False})
+    assert registry.tool_defs(db, db.get(User, user_id)) == []
+
+
+def test_a_tool_absent_from_the_overrides_is_on(db, user_id):
+    """Absent means on, the rule the model capability flags follow."""
+    from lembas.db.models import User
+
+    _server(db, tool_overrides_json={"something_else": False})
+    assert len(registry.tool_defs(db, db.get(User, user_id))) == 1
+
+
+async def test_refresh_caches_the_tools_and_clears_the_last_error(db, mock_http):
+    mock_http(FakeServer())
+    server = _server(db, tools_json=[], last_error="it was broken")
+
+    count, error = await registry.refresh(db, server)
+    assert (count, error) == (1, "")
+    db.refresh(server)
+    assert server.tools_json[0]["offer_name"] == "github_search"
+    assert server.last_error == ""
+    assert server.protocol_version == "2025-06-18"
+    assert server.server_info["name"] == "fake"
+
+
+async def test_a_failed_refresh_records_it_and_keeps_the_cached_tools(db, mock_http):
+    mock_http(lambda _r: httpx.Response(500, text="down"))
+    server = _server(db)
+
+    count, error = await registry.refresh(db, server)
+    assert count == 0 and "500" in error
+    db.refresh(server)
+    assert server.last_error
+    assert server.last_checked_at is not None
+    assert len(server.tools_json) == 1, "a bad refresh must not empty the list"
+
+
+async def test_a_refresh_keeps_a_choice_about_a_tool_that_survives(db, mock_http):
+    mock_http(FakeServer(tools=[{"name": "search"}, {"name": "gone"}]))
+    server = _server(db, tool_overrides_json={"search": False, "vanished": False})
+    await registry.refresh(db, server)
+
+    db.refresh(server)
+    assert server.tool_overrides_json == {"search": False}
+
+
+async def test_a_failing_tool_call_reaches_the_model_as_words(db, user_id, mock_http):
+    from lembas.db.models import User
+
+    mock_http(lambda _r: httpx.Response(500, text="down"))
+    _server(db)
+    tool = registry.tool_defs(db, db.get(User, user_id))[0]
+
+    outcome = await tool.run(None, {})
+    assert outcome.event["status"] == "error"
+    assert "GitHub" in outcome.content
+
+
+def test_two_servers_exposing_the_same_tool_do_not_collide(db, user_id):
+    from lembas.db.models import User
+
+    _server(db)
+    _server(db, slug="gitlab", name="GitLab", tools_json=[{"name": "search", "schema": {}}])
+    names = [tool.name for tool in registry.tool_defs(db, db.get(User, user_id))]
+    assert len(names) == len(set(names)) == 2
+
+
+def test_a_disabled_server_offers_nothing(db, user_id):
+    from lembas.db.models import User
+
+    _server(db, enabled=False)
+    assert registry.tool_defs(db, db.get(User, user_id)) == []
+    assert db.scalar(select(McpServer)) is not None