Documentation
MCP Servers
The Model Context Protocol (MCP) allows you to connect your AI agents to external tools and data sources. Extend GRID's capabilities with filesystem access, database connections, and more.
Adding an MCP Server
GRID supports two types of MCP connections:
1. stdio (Local/Container)
Runs a command within the environment to start the MCP server. Perfect for local tools and containerized environments.
Command
The executable (e.g.,
npx, python)Arguments
Arguments to pass (e.g.,
-y @modelcontextprotocol/server-filesystem)Environment Variables
Optional env vars needed by the server
2. HTTP (Remote)
Connects to an MCP server running on a remote URL via Server-Sent Events (SSE). Ideal for cloud services and distributed systems.
URL
The full endpoint URL (e.g.,
https://api.myapp.com/mcp/sse)Common MCP Servers
| Server | Description | Example Command |
|---|---|---|
| Filesystem | Read/write files | npx -y @modelcontextprotocol/server-filesystem |
| PostgreSQL | Database access | npx -y @modelcontextprotocol/server-postgres |
| GitHub | Repo management | npx -y @modelcontextprotocol/server-github |
Using MCP Tools
Once configured, MCP tools are automatically available to your Custom Agents. You can enable/disable specific servers for different agents in the agent configuration.
MCP tools appear automatically in agent tool selection
Enable/disable per agent for fine-grained control
Tools are scoped to the workspace environment
Full access to server capabilities via protocol