The Trustgrid MCP server lets AI assistants — Claude, Copilot, Cursor, and others — interact with your Trustgrid environment using natural language. Ask questions about node status, run network diagnostics, query alerts, and explore your topology without leaving your editor or chat interface.

What you can do#

⚠️ Read-only access
The MCP server provides read-only access to your Trustgrid infrastructure. No changes can be made through the MCP interface at this time.

Endpoint URL#

The MCP server is hosted at https://mcp.<domain>.trustgrid.io where <domain> matches your organization's Trustgrid domain. For example, if your portal is at acme.trustgrid.io, your MCP endpoint is https://mcp.acme.trustgrid.io.

The default production endpoint is:

https://mcp.trustgrid.io/mcp/<group>/[<group>...]

The server exposes three tool groups, each at its own path:

Path Group Purpose
/mcp or /mcp/codemode codemode AI sandbox with full read API access
/mcp/read read Direct read-only API tools
/mcp/tools tools Live node diagnostic tools
/mcp/codemode/read codemode + read Both tool sets combined
/mcp/all all Everything

You can put any number of suffixes after /mcp to combine groups as needed. For example, /mcp/codemode/tools gives you both codemode and traditional MCP tools access.

See Tools for details on what each group provides.

Transport#

The Trustgrid MCP server uses Streamable HTTP transport. This is the only transport available for external integrations — there is no stdio or WebSocket option. Your MCP client must support HTTP-based MCP connections.

Authentication#

The server accepts two authentication types:

See Authentication for setup details.

Rate limits#

Requests are rate-limited per request:

Group Limit
codemode 10 requests / 60 seconds
read 30 requests / 60 seconds
tools 10 requests / 60 seconds

The server returns 429 Too Many Requests with Retry-After and X-RateLimit-* headers when limits are exceeded.

Next steps#