Mcp by Matih Labs is an MCP server published under the package @matihlabs/mcp (version 0.2.1) on npm. It is described as a grounded data layer for LLMs, covering governed SQL execution, metrics, data lineage, and a data catalog over your existing data infrastructure. The server communicates over stdio transport and requires a single environment variable, MATIH_MCP_TOKEN, for authentication. The repository is hosted at github.com/matih-labs/matih-mcp.
The core proposition is governance: rather than allowing an LLM to run arbitrary queries, this server sits as an intermediary that applies controls and structure — governed SQL, tracked lineage, defined metrics, and a browsable catalog. This makes it relevant for organizations that need auditability or policy enforcement around how AI agents interact with their data.
No tool definitions were provided in the registry record, so the specific callable operations exposed to an agent cannot be enumerated here. Developers evaluating this server should review the upstream repository to understand exactly what tools are available before integrating it into an agent workflow. The server requires Node.js tooling to install from npm and a valid MATIH_MCP_TOKEN credential to operate.