Agent: tool profiles and profile-scoped MCP connections

- Extend tool profile helpers for MCP server key resolution and filtering
- Lazily connect/disconnect MCP servers per active profile in AgentLoop
- Harden MCP client (timeouts, tool naming, connect_mcp_server entry)
- Adjust context and tool modules to align with profile-aware tooling
- docker-compose: minor gateway/workspace notes

Made-with: Cursor

docker-compose.yml

@@ -4,6 +4,8 @@ x-common-config: &common-config
     dockerfile: Dockerfile
   volumes:
     - ~/.nanobot:/root/.nanobot
+    # Host repo ./workspace → /workspace in container. Set agents.defaults.workspace to /workspace.
+    - ./workspace:/workspace
 
 services:
   nanobot-gateway:

nanobot/agent/context.py

@@ -101,6 +101,12 @@ Your workspace is at: {workspace_path}
 - History log: {workspace_path}/memory/HISTORY.md (grep-searchable)
 - Custom skills: {workspace_path}/skills/{{skill-name}}/SKILL.md
 
+**Filesystem tools (read_file, write_file, edit_file, list_dir):** Use paths **under this workspace root only** (`{workspace_path}`). Do not invent other roots (e.g. `/mnt/data/...` on a host) unless you know they are valid on this runtime. **`list_dir` takes one directory path**—no wildcards (never pass `*.pdf` in the path). To find PDFs, `list_dir("{workspace_path}")` (or a subfolder) and filter for `.pdf` names, or use `exec` with `find` under that directory.
+
+**Answering after tools:** When a tool already returned what the user needs, base your reply **only on that tool output**—same topic as the user’s question, no hijacking.
+- After **`list_dir`:** If they asked for PDFs (or another extension), list **only** matching names (paths under `{workspace_path}` if useful). If none, say so briefly. No essays, no calling the folder "code" unless they asked for analysis.
+- After **`read_emails`:** Answer **only** from the email text the tool returned (From, Subject, Date, attachments, downloaded paths, body as needed). Do **not** switch to unrelated topics (Git, Gitea, this repo, workspace docs, coding help, general chit-chat). Do **not** apologize at length or describe "what an email is". Match the question: e.g. “latest email” → sender + subject (+ date) in a few lines unless they asked for the full body.
+
 ## Gitea API (This Repository)
 **CRITICAL**: This repository uses Gitea at `http://10.0.30.169:3000/api/v1`, NOT GitHub.
 - Repository: `ilia/nanobot`
@@ -120,7 +126,7 @@ Always be helpful, accurate, and concise. Before calling tools, briefly tell the
 When remembering something important, write to {workspace_path}/memory/MEMORY.md
 To recall past events, grep {workspace_path}/memory/HISTORY.md
 
-IMPORTANT: For email queries (latest email, email sender, inbox, etc.), ALWAYS use the read_emails tool. NEVER use exec() with mail/tail/awk commands or read_file() on /var/mail - those will not work. The read_emails tool is the only way to access emails."""
+IMPORTANT: For email queries (latest email, email sender, inbox, etc.), ALWAYS use the read_emails tool. NEVER use exec() with mail/tail/awk commands or read_file() on /var/mail - those will not work. The read_emails tool is the only way to access emails. Once read_emails returns, your assistant reply must **only** satisfy that email question from the tool result—ignore Gitea/workspace/bootstrap content unless the user tied their question to it."""
 
     def _load_bootstrap_files(self) -> str:
         """Load all bootstrap files from workspace."""

nanobot/agent/loop.py

@@ -90,8 +90,8 @@ class AgentLoop:
 
         self._running = False
         self._mcp_servers = mcp_servers or {}
-        self._mcp_stack: AsyncExitStack | None = None
-        self._mcp_connected = False
+        self._mcp_stacks: dict[str, AsyncExitStack] = {}
+        self._mcp_connected_servers: set[str] = set()
         self._tool_profiles: dict = tool_profiles or {}
         self._default_tool_profile = default_tool_profile
         self._tool_routing = tool_routing or ToolRoutingConfig()
@@ -135,7 +135,10 @@ class AgentLoop:
             from nanobot.config.loader import load_config
             config = load_config()
             if config.channels.email.enabled:
-                email_tool = EmailTool(email_config=config.channels.email)
+                email_tool = EmailTool(
+                    email_config=config.channels.email,
+                    workspace=self.workspace,
+                )
                 self.tools.register(email_tool)
                 logger.info(f"Email tool '{email_tool.name}' registered successfully")
             else:
@@ -159,15 +162,69 @@ class AgentLoop:
             logger.warning(f"Calendar tool not available: {e}")
             # Calendar tool not available or not configured - silently skip
 
-    async def _connect_mcp(self) -> None:
-        """Connect to configured MCP servers (one-time, lazy)."""
-        if self._mcp_connected or not self._mcp_servers:
+    def _unregister_mcp_tools_for_server(self, server_key: str) -> None:
+        """Remove tools registered from one MCP server (prefix mcp_<key>_)."""
+        prefix = f"mcp_{server_key}_"
+        for name in list(self.tools.tool_names):
+            if name.startswith(prefix):
+                self.tools.unregister(name)
+
+    async def _disconnect_mcp_server(self, server_key: str) -> None:
+        """Close one MCP server and remove its tools (used when switching tool profiles)."""
+        stack = self._mcp_stacks.pop(server_key, None)
+        if stack is not None:
+            try:
+                await stack.aclose()
+            except (RuntimeError, BaseExceptionGroup):
+                pass
+        self._unregister_mcp_tools_for_server(server_key)
+        self._mcp_connected_servers.discard(server_key)
+        logger.info(f"MCP server '{server_key}': disconnected")
+
+    async def _sync_mcp_to_profile_needs(self, needed_keys: list[str]) -> None:
+        """
+        Ensure only MCP servers in needed_keys are connected: tear down extras, connect missing.
+
+        When tools.toolProfiles is empty, pass the full configured key list so all servers stay up.
+        """
+        if not self._mcp_servers:
             return
-        self._mcp_connected = True
-        from nanobot.agent.tools.mcp import connect_mcp_servers
-        self._mcp_stack = AsyncExitStack()
-        await self._mcp_stack.__aenter__()
-        await connect_mcp_servers(self._mcp_servers, self.tools, self._mcp_stack)
+        needed = set(needed_keys)
+        for key in list(self._mcp_connected_servers):
+            if key not in needed:
+                await self._disconnect_mcp_server(key)
+        connect_order = [k for k in self._mcp_servers.keys() if k in needed]
+        await self._ensure_mcp_servers_connected(connect_order)
+
+    async def _ensure_mcp_servers_connected(self, server_keys: list[str]) -> None:
+        """Lazily connect MCP servers (each gets its own AsyncExitStack for per-server teardown)."""
+        if not self._mcp_servers or not server_keys:
+            return
+        pending = [
+            k
+            for k in server_keys
+            if k in self._mcp_servers and k not in self._mcp_connected_servers
+        ]
+        if not pending:
+            return
+
+        from nanobot.agent.tools.mcp import connect_mcp_server
+
+        for key in pending:
+            stack = AsyncExitStack()
+            await stack.__aenter__()
+            try:
+                await connect_mcp_server(
+                    key, self._mcp_servers[key], self.tools, stack
+                )
+                self._mcp_stacks[key] = stack
+                self._mcp_connected_servers.add(key)
+            except Exception as e:
+                logger.error(f"MCP server '{key}': failed to connect: {e}")
+                try:
+                    await stack.aclose()
+                except (RuntimeError, BaseExceptionGroup):
+                    pass
 
     def _set_tool_context(self, channel: str, chat_id: str) -> None:
         """Update context for all tools that need routing info."""
@@ -255,26 +312,37 @@ class AgentLoop:
         final_content = None
         tools_used: list[str] = []
 
-        from nanobot.agent.tool_profiles import compute_allowed_tool_names
+        from nanobot.agent.tool_profiles import (
+            compute_allowed_tool_names,
+            mcp_keys_to_connect,
+        )
         from nanobot.agent.tool_routing import is_tool_not_found_error
 
-        tools_full = self.tools.get_definitions()
+        configured_mcp = list(self._mcp_servers.keys())
         tools_expanded = False
         allowed_names: set[str] | None = None
+
         if self._tool_profiles:
             routing_text = self._extract_routing_text(initial_messages)
             profile_key = await self._pick_tool_profile(routing_text)
             prof = self._tool_profiles[profile_key]
+            await self._sync_mcp_to_profile_needs(
+                mcp_keys_to_connect(prof, configured_mcp)
+            )
             always = set(self._tool_routing.always_include_tools)
             allowed_names = compute_allowed_tool_names(
                 self.tools,
                 prof,
-                list(self._mcp_servers.keys()),
+                configured_mcp,
                 always,
             )
             logger.info(
                 f"Tool profile '{profile_key}': {len(allowed_names)}/{len(self.tools)} tools exposed"
             )
+        else:
+            await self._sync_mcp_to_profile_needs(configured_mcp)
+
+        tools_full = self.tools.get_definitions()
 
         while iteration < self.max_iterations:
             iteration += 1
@@ -338,6 +406,8 @@ class AgentLoop:
                         and is_tool_not_found_error(result)
                     ):
                         tools_expanded = True
+                        await self._sync_mcp_to_profile_needs(configured_mcp)
+                        tools_full = self.tools.get_definitions()
                         logger.info(
                             "Expanded tool set to full registry (missing tool after profile filter)"
                         )
@@ -358,7 +428,6 @@ class AgentLoop:
     async def run(self) -> None:
         """Run the agent loop, processing messages from the bus."""
         self._running = True
-        await self._connect_mcp()
         logger.info("Agent loop started")
 
         while self._running:
@@ -382,13 +451,13 @@ class AgentLoop:
                 continue
 
     async def close_mcp(self) -> None:
-        """Close MCP connections."""
-        if self._mcp_stack:
-            try:
-                await self._mcp_stack.aclose()
-            except (RuntimeError, BaseExceptionGroup):
-                pass  # MCP SDK cancel scope cleanup is noisy but harmless
-            self._mcp_stack = None
+        """Close all MCP connections and drop MCP tools from the registry."""
+        for key in list(
+            set(self._mcp_stacks.keys()) | self._mcp_connected_servers
+        ):
+            await self._disconnect_mcp_server(key)
+        self._mcp_stacks.clear()
+        self._mcp_connected_servers.clear()
 
     def stop(self) -> None:
         """Stop the agent loop."""
@@ -658,7 +727,6 @@ Respond with ONLY valid JSON, no markdown fences."""
         Returns:
             The agent's response.
         """
-        await self._connect_mcp()
         msg = InboundMessage(
             channel=channel,
             sender_id="user",

nanobot/agent/tool_profiles.py

@@ -4,9 +4,12 @@ from __future__ import annotations
 
 from typing import TYPE_CHECKING
 
+from loguru import logger
+
+from nanobot.config.schema import ToolProfileConfig
+
 if TYPE_CHECKING:
     from nanobot.agent.tools.registry import ToolRegistry
-    from nanobot.config.schema import ToolProfileConfig
 
 
 def mcp_server_for_tool(tool_name: str, mcp_server_keys: list[str]) -> str | None:
@@ -26,6 +29,32 @@ def mcp_server_for_tool(tool_name: str, mcp_server_keys: list[str]) -> str | Non
     return None
 
 
+def mcp_keys_to_connect(
+    profile: ToolProfileConfig, configured_mcp_keys: list[str]
+) -> list[str]:
+    """
+    Config keys for MCP servers to connect for this profile, in config order.
+
+    None on profile.mcp_servers means all configured servers; [] means none.
+    Unknown keys in the profile list are logged and skipped.
+    """
+    if not configured_mcp_keys:
+        return []
+    configured_set = set(configured_mcp_keys)
+    if profile.mcp_servers is None:
+        return list(configured_mcp_keys)
+    out: list[str] = []
+    for k in profile.mcp_servers:
+        if k in configured_set:
+            out.append(k)
+        else:
+            logger.warning(
+                f"tools.toolProfiles entry references unknown MCP server {k!r}; "
+                "not in tools.mcpServers keys"
+            )
+    return out
+
+
 def compute_allowed_tool_names(
     registry: ToolRegistry,
     profile: ToolProfileConfig,

nanobot/agent/tools/email.py

@@ -3,6 +3,7 @@
 import asyncio
 import imaplib
 import ssl
+from pathlib import Path
 from datetime import date
 from email import policy
 from email.header import decode_header, make_header
@@ -36,17 +37,20 @@ class EmailTool(Tool):
         "unread_only (bool, default false), mark_seen (bool, default false), download_attachments (bool, default false "
         "- set to true to download all attachments to workspace), attachment_name (string, optional - filter emails by "
         "attachment filename, case-insensitive partial match). Returns formatted email list with sender, subject, date, "
-        "attachments (if any), downloaded file paths (if downloaded), and body."
+        "attachments (if any), downloaded file paths (if downloaded), and body. After you receive this output, your "
+        "reply to the user must address their email question using only this data—no unrelated topics."
     )
     
-    def __init__(self, email_config: Any = None):
+    def __init__(self, email_config: Any = None, workspace: Path | None = None):
         """
         Initialize email tool with email configuration.
         
         Args:
             email_config: Optional EmailConfig instance. If None, loads from config.
+            workspace: Directory for downloaded attachments (defaults to config workspace_path).
         """
         self._email_config = email_config
+        self._workspace = workspace
     
     @property
     def config(self) -> Any:
@@ -315,8 +319,12 @@ class EmailTool(Tool):
                 if download_attachments and attachments:
                     import logging
                     logger = logging.getLogger(__name__)
-                    from pathlib import Path
-                    workspace = Path("/mnt/data/nanobot/workspace")
+                    if self._workspace is not None:
+                        workspace = Path(self._workspace).expanduser().resolve()
+                    else:
+                        from nanobot.config.loader import load_config
+
+                        workspace = load_config().workspace_path.expanduser().resolve()
                     workspace.mkdir(parents=True, exist_ok=True)
                     
                     # Build a map of attachment parts by decoded filename for efficient lookup

nanobot/agent/tools/filesystem.py

@@ -28,7 +28,9 @@ class ReadFileTool(Tool):
     
     @property
     def description(self) -> str:
-        return """Read the contents of a file at the given path. 
+        return """Read the contents of a file at the given path.
+
+`path` must be a single file path under the configured workspace (no `*` globs).
         
 ALWAYS use this tool to read files - it supports:
 - Text files (plain text, code, markdown, etc.)
@@ -44,7 +46,7 @@ For reading files, use read_file FIRST. Only use exec for complex data processin
             "properties": {
                 "path": {
                     "type": "string",
-                    "description": "The file path to read"
+                    "description": "Absolute or workspace-relative path to one file (no wildcards)",
                 }
             },
             "required": ["path"]
@@ -115,7 +117,7 @@ class WriteFileTool(Tool):
     
     @property
     def description(self) -> str:
-        return "Write content to a file at the given path. Creates parent directories if needed. IMPORTANT: Always provide both 'path' and 'content' parameters. If no full path is specified, use the workspace directory (/mnt/data/nanobot/workspace/)."
+        return "Write content to a file at the given path. Creates parent directories if needed. IMPORTANT: Always provide both 'path' and 'content' parameters. Paths must be under the workspace root from the system prompt (no globs)."
     
     @property
     def parameters(self) -> dict[str, Any]:
@@ -219,7 +221,11 @@ class ListDirTool(Tool):
     
     @property
     def description(self) -> str:
-        return "List the contents of a directory."
+        return (
+            "List files and subfolders in one directory. "
+            "`path` must be a directory that exists under the workspace root—no `*` or `*.pdf` wildcards. "
+            "To list PDFs, list the directory and read names ending in .pdf, or use exec with find."
+        )
     
     @property
     def parameters(self) -> dict[str, Any]:
@@ -228,7 +234,7 @@ class ListDirTool(Tool):
             "properties": {
                 "path": {
                     "type": "string",
-                    "description": "The directory path to list"
+                    "description": "Path to an existing directory under the workspace (no wildcards)",
                 }
             },
             "required": ["path"]

nanobot/agent/tools/mcp.py

@@ -1,5 +1,8 @@
 """MCP client: connects to MCP servers and wraps their tools as native nanobot tools."""
 
+import asyncio
+import json
+import re
 from contextlib import AsyncExitStack
 from typing import Any
 
@@ -9,15 +12,65 @@ from nanobot.agent.tools.base import Tool
 from nanobot.agent.tools.registry import ToolRegistry
 
 
+_SAFE_TOOL_NAME_RE = re.compile(r"[^A-Za-z0-9_]+")
+
+
+def _normalize_tool_segment(segment: str) -> str:
+    """
+    Normalize MCP server/tool names into a safe function name segment.
+
+    - Replace non [A-Za-z0-9_] with underscore
+    - Collapse repeated underscores
+    - Trim leading/trailing underscores
+    - Ensure non-empty
+    """
+    s = _SAFE_TOOL_NAME_RE.sub("_", (segment or "").strip())
+    s = re.sub(r"_+", "_", s).strip("_")
+    return s or "tool"
+
+
+def _render_mcp_content_blocks(blocks: list[Any]) -> str:
+    """Render MCP content blocks into a stable, readable string."""
+    from mcp import types
+
+    parts: list[str] = []
+    for block in blocks or []:
+        if isinstance(block, types.TextContent):
+            parts.append(block.text)
+            continue
+
+        # Prefer structured JSON for non-text blocks when possible.
+        dump = getattr(block, "model_dump", None)
+        if callable(dump):
+            try:
+                parts.append(json.dumps(dump(), ensure_ascii=False, indent=2))
+                continue
+            except Exception:
+                pass
+
+        parts.append(str(block))
+    return "\n".join([p for p in parts if p is not None]).strip()
+
+
 class MCPToolWrapper(Tool):
     """Wraps a single MCP server tool as a nanobot Tool."""
 
-    def __init__(self, session, server_name: str, tool_def):
+    def __init__(
+        self,
+        session,
+        *,
+        server_key: str,
+        tool_def,
+        registered_name: str,
+        call_timeout_s: float = 30.0,
+    ):
         self._session = session
         self._original_name = tool_def.name
-        self._name = f"mcp_{server_name}_{tool_def.name}"
+        self._server_key = server_key
+        self._name = registered_name
         self._description = tool_def.description or tool_def.name
         self._parameters = tool_def.inputSchema or {"type": "object", "properties": {}}
+        self._call_timeout_s = call_timeout_s
 
     @property
     def name(self) -> str:
@@ -32,71 +85,86 @@ class MCPToolWrapper(Tool):
         return self._parameters
 
     async def execute(self, **kwargs: Any) -> str:
-        from mcp import types
-        import json
-        result = await self._session.call_tool(self._original_name, arguments=kwargs)
-        parts = []
-        for block in result.content:
-            if isinstance(block, types.TextContent):
-                parts.append(block.text)
-            else:
-                parts.append(str(block))
-        output = "\n".join(parts)
-        
-        # For empty results from search/list operations, provide clearer feedback
-        if not output or output.strip() == "":
-            # Check if this is a search/list operation (common patterns)
-            if "search" in self._original_name.lower() or "list" in self._original_name.lower():
-                if "unread" in str(kwargs).lower() or "is:unread" in str(kwargs).lower():
-                    return "No unread emails found."
-                return "No results found."
-        
-        # Try to parse JSON to check for empty arrays/lists
+        try:
+            result = await asyncio.wait_for(
+                self._session.call_tool(self._original_name, arguments=kwargs),
+                timeout=self._call_timeout_s,
+            )
+        except asyncio.TimeoutError:
+            return (
+                f"Error: MCP tool timed out after {self._call_timeout_s:.0f}s "
+                f"({self._server_key}:{self._original_name})"
+            )
+
+        output = _render_mcp_content_blocks(getattr(result, "content", []))
+        if not output:
+            return "(no output)"
+
+        # If the tool returned JSON, normalize empty collections to a clearer message.
         try:
             parsed = json.loads(output)
-            if isinstance(parsed, list) and len(parsed) == 0:
-                if "search" in self._original_name.lower() or "list" in self._original_name.lower():
-                    if "unread" in str(kwargs).lower() or "is:unread" in str(kwargs).lower():
-                        return "No unread emails found."
-                    return "No results found."
+            if parsed == [] or parsed == {}:
+                return "No results found."
         except (json.JSONDecodeError, ValueError):
             pass  # Not JSON, continue with original output
-        
-        return output or "(no output)"
+
+        return output
+
+
+async def connect_mcp_server(
+    name: str, cfg: Any, registry: ToolRegistry, stack: AsyncExitStack
+) -> None:
+    """Connect one MCP server and register its tools (used for lazy profile-scoped connections)."""
+    from mcp import ClientSession, StdioServerParameters
+    from mcp.client.stdio import stdio_client
+
+    if cfg.command:
+        params = StdioServerParameters(
+            command=cfg.command, args=cfg.args, env=cfg.env or None
+        )
+        read, write = await stack.enter_async_context(stdio_client(params))
+    elif cfg.url:
+        from mcp.client.streamable_http import streamable_http_client
+
+        read, write, _ = await stack.enter_async_context(
+            streamable_http_client(cfg.url)
+        )
+    else:
+        logger.warning(f"MCP server '{name}': no command or url configured, skipping")
+        return
+
+    session = await stack.enter_async_context(ClientSession(read, write))
+    await session.initialize()
+
+    tools = await session.list_tools()
+    for tool_def in tools.tools:
+        safe_server = _normalize_tool_segment(name)
+        safe_tool = _normalize_tool_segment(tool_def.name)
+        base = f"mcp_{safe_server}_{safe_tool}"
+        registered_name = base
+        i = 2
+        while registry.has(registered_name):
+            registered_name = f"{base}_{i}"
+            i += 1
+
+        wrapper = MCPToolWrapper(
+            session,
+            server_key=name,
+            tool_def=tool_def,
+            registered_name=registered_name,
+        )
+        registry.register(wrapper)
+        logger.debug(f"MCP: registered tool '{wrapper.name}' from server '{name}'")
+
+    logger.info(f"MCP server '{name}': connected, {len(tools.tools)} tools registered")
 
 
 async def connect_mcp_servers(
     mcp_servers: dict, registry: ToolRegistry, stack: AsyncExitStack
 ) -> None:
-    """Connect to configured MCP servers and register their tools."""
-    from mcp import ClientSession, StdioServerParameters
-    from mcp.client.stdio import stdio_client
-
+    """Connect to every configured MCP server and register their tools."""
     for name, cfg in mcp_servers.items():
         try:
-            if cfg.command:
-                params = StdioServerParameters(
-                    command=cfg.command, args=cfg.args, env=cfg.env or None
-                )
-                read, write = await stack.enter_async_context(stdio_client(params))
-            elif cfg.url:
-                from mcp.client.streamable_http import streamable_http_client
-                read, write, _ = await stack.enter_async_context(
-                    streamable_http_client(cfg.url)
-                )
-            else:
-                logger.warning(f"MCP server '{name}': no command or url configured, skipping")
-                continue
-
-            session = await stack.enter_async_context(ClientSession(read, write))
-            await session.initialize()
-
-            tools = await session.list_tools()
-            for tool_def in tools.tools:
-                wrapper = MCPToolWrapper(session, name, tool_def)
-                registry.register(wrapper)
-                logger.debug(f"MCP: registered tool '{wrapper.name}' from server '{name}'")
-
-            logger.info(f"MCP server '{name}': connected, {len(tools.tools)} tools registered")
+            await connect_mcp_server(name, cfg, registry, stack)
         except Exception as e:
             logger.error(f"MCP server '{name}': failed to connect: {e}")