← Documentation and encyclopedia

Developers and engineers

MCP CLI and baseline comparison

Inspect a trusted server from a repository checkout and interpret the result.

Reviewed

Source-workspace setup

@pentriarch/mcp-scan is currently a private repository package. These commands require an authorised checkout with workspace dependencies installed; they do not describe a public npm installation.

Use the repository’s documented Node and pnpm requirements. Only launch trusted code in stdio mode: static inspection avoids tool calls but still starts the server process.

Remote inspection

Use an endpoint you are permitted to assess. --http selects streamable HTTP; --sse selects SSE.

pnpm --filter @pentriarch/mcp-scan scan --http https://your-owned-host.example/mcp --json
pnpm --filter @pentriarch/mcp-scan scan --sse https://your-owned-host.example/mcp

Baseline and comparison

Save and review a known baseline, then compare it after a change. A baseline records an observation; it is not an approval certificate.

pnpm --filter @pentriarch/mcp-scan scan --stdio "node ./my-server.mjs" --save-baseline baseline.json
pnpm --filter @pentriarch/mcp-scan scan --stdio "node ./my-server.mjs" --baseline baseline.json --fail-on medium

Automation and exit status

Exit 0 means below the configured threshold, exit 1 means findings meet or exceed it, and exit 2 means an error. The default threshold is high. An error is not a clean assessment.

Retain machine-readable output. Distinguish a threshold failure from a connection failure. Do not automatically replace an unreviewed baseline merely to pass a pipeline.

Continue reading

Need help with availability or an assessment? Contact Pentriarch support.