About the Query MCP Server
Access the Query Security Data Mesh from any MCP client
TL;DRThe Query MCP Server exposes the Query Security Data Mesh over the Model Context Protocol. It is remote — nothing to install or run locally.
Connect once and a client reaches every source on the mesh. New sources are available with no change to the client.
Authenticate with a Query API token.
What It Exposes
| Group | Tools |
|---|---|
| Search and discovery | Execute, Validate, and generate FSQL; search the schema; list connectors |
| Detection translation | KQL to FSQL, Sigma to FSQL, FSQL to Query recipe |
| Query Workers | Launch an investigation, list and read its artifacts |
| Session state | Create, read, and update session artifacts |
| Organization context | Read environment profile and org policies; update environment profile |
All tools read except Update Environment Profile. See the MCP Tool Reference.
How It Works
Tool calls become federated searches that execute against your connected sources where the data already lives. Schema is normalized to OCSF at query time, and results are returned to the calling client. Nothing is ingested or copied.
Connecting
| Setting | Value |
|---|---|
| URL | https://mcp.query.ai/mcp |
| Transport | Streamable HTTP |
Two authentication methods are supported:
- OAuth 2.0 with dynamic client registration. The client registers itself and you sign in through the browser. No credential is stored in the client's configuration.
- API token. A Query API token sent in the
x-token-authorizationheader, for clients that do not support OAuth.
See Connecting an MCP Client for client configuration, or Connecting Charlotte AI to the Query Mesh for AgentWorks.
An API token grants access to every source on your mesh. Store it in your client's secret store and rotate it if exposed. OAuth avoids holding a long-lived credential in client configuration.
What's Next
Updated about 1 hour ago