Use MCP Tools

Model Context Protocol (MCP) servers expose additional tools to Snippbot agents. Once connected, tools from MCP servers appear alongside built-in tools and can be called by any agent.

What is MCP?

MCP is an open protocol that lets AI agents communicate with external tool servers using a standard interface. An MCP server exposes a set of tools (functions the agent can call) over either:

• stdio — The agent launches the server as a subprocess and communicates via stdin/stdout
• HTTP/SSE — The agent connects to an already-running HTTP server

Built-in catalog

Expand

Snippbot ships with 21 pre-built MCP server entries ready to install:

Category	Servers
Productivity & Knowledge	Google Drive, Microsoft 365, Notion, Slack, Obsidian
Development & DevOps	GitHub, Filesystem, PostgreSQL, Supabase, Kubernetes
Web & Automation	Brave Search, DuckDuckGo Search, Playwright Browser, Puppeteer Browser, Zapier MCP
Specialized Tools	Figma, Stripe, Blender MCP
Quality & Intelligence	Context7, Sequential Thinking, MCP Memory

Connecting an MCP server

1. Open the Skills page in the sidebar

2. Click “Add Server” or browse the Catalog tab

3. Configure the server — fill in the form fields:

   • Name — a label for the server (e.g., “my-filesystem”)
   • Transport — select stdio or HTTP/SSE
   • Command — the executable to launch (e.g., npx)
   • Args — command arguments (e.g., -y @modelcontextprotocol/server-filesystem /workspace)

4. Click “Connect” — Snippbot launches the server, discovers its tools, and displays them

5. Enable the server for agents that should have access

Catalog install (one click)

Expand

From the Catalog tab, click any pre-built server. A configuration dialog appears where you fill in any required fields (such as a personal access token for GitHub). Click Install and the server is configured, connected, and tools discovered automatically.

Tool namespacing

MCP tools appear in agents with namespaced names to avoid conflicts:

Tools follow the naming format mcp__{server_name}__{tool_name} — for example, mcp__github__create_issue, mcp__filesystem__read_file, or mcp__postgresql__execute_query.

Agents see each tool’s description prefixed with the server name, such as “[GitHub] Create a new issue in a repository”.

Transport configuration

When adding a server on the Skills page, you choose the transport type and fill in the relevant fields.

stdio transport

For servers that run as a local subprocess. Fill in:

Field	Description	Example
Name	Server label	filesystem
Transport	Select stdio	—
Command	Executable to launch	npx
Args	Command arguments	-y @modelcontextprotocol/server-filesystem /home/user/projects

The server process is started on demand and kept alive for the session.

HTTP/SSE transport

For servers running as a service. Fill in:

Field	Description	Example
Name	Server label	my-api-server
Transport	Select HTTP/SSE	—
URI	Server URL	http://localhost:3000/mcp

OAuth2 authentication for HTTP servers

For HTTP MCP servers that require OAuth2, expand the Authentication section in the server form and fill in:

Field	Description
Auth Type	Select OAuth2 Client Credentials
Token URL	The provider’s token endpoint (e.g., https://auth.example.com/oauth2/token)
Client ID	Your OAuth2 client ID
Client Secret	Your OAuth2 client secret
Scope	Required scopes (e.g., mcp:read mcp:write)

Snippbot handles token acquisition, caching, and refresh automatically. Tokens are refreshed 60 seconds before expiry.

Testing a connection

From the Skills page, select a server and click Test Connection. The result displays connection status, the number of discovered tools, latency, and server capabilities.

Managing servers

Manage MCP servers from the Skills page in the sidebar. You can connect, disconnect, view discovered tools, and remove servers directly from the UI. Each server card shows its connection status, discovered tools count, and provides actions to connect, disconnect, or remove.

Assigning tools to agents

By default, all enabled MCP servers’ tools are available to all agents. To restrict:

1. Open the agent detail page → Settings tab

2. Under Tools, deselect specific MCP tools or entire servers

3. Save

When spawning a sub-agent, the parent agent can also restrict tool access. See Use Sub-Agents for details on tool isolation.

Creating your own MCP server

Any HTTP or stdio server implementing the MCP protocol can be connected. A minimal Python example:

from mcp.server import Server
from mcp.types import Tool, TextContent

app = Server("my-tools")

@app.list_tools()
async def list_tools():
    return [Tool(name="greet", description="Greet someone", inputSchema={
        "type": "object",
        "properties": {"name": {"type": "string"}},
        "required": ["name"]
    })]

@app.call_tool()
async def call_tool(name, arguments):
    if name == "greet":
        return [TextContent(type="text", text=f"Hello, {arguments['name']}!")]

Run it: python server.py and connect with transport: stdio, command: python, args: [server.py].

Tip

For a list of community-built MCP servers, see modelcontextprotocol.io/servers.