MCP channel

Argus exposes an MCP (Model Context Protocol) server so that other AI tools — Claude Code, Claude Desktop, Cursor, your own agents — can consult it as a colleague, not a toolbox. The external AI delegates a question or a review; Argus runs its own org-aware loop with its own scanners and returns the result. Your organization’s knowledge — its identity (SOUL), the summary Argus keeps across sessions (MEMORY), and your context documents — never leaves Argus.

Setup

The channel is served on the HTTP front door at /mcp. Enable it and mint a token:

argus.yaml

channels:
  mcp:
    type: mcp

argus user add alice
argus user mcp-token create alice    # prints the cleartext once, stores only its hash

Point your MCP client at the endpoint with the token as a bearer credential:

URL:            https://argus.your-company.example/mcp
Authorization:  Bearer <token>

Every call is attributed to the Person who owns the token — the human Argus recognizes behind the identity mcp:<token-hash> — with their role and audit trail. Revoke anytime with argus user mcp-token revoke.

What’s exposed

The surface is deliberately small — a few coarse capabilities, never the low-level scanner tools:

• review — a security review. Two targets:
  • snapshot: the caller supplies the files to review;
  • repo: a repo + ref on a connected code host, which Argus checks out and reviews in full.
• consult — ask a question against the organization’s knowledge (SOUL, curated context, recent findings).
• Resources — read-only org knowledge: SOUL, your context documents (~/.argus/context/), recent reports.

What it’s not

Generic security Q&A (“what is path traversal?”) is a non-goal — your AI tool already answers that on its own. The MCP surface exists for what only Argus can do: reviews with real scanners, grounded in your organization’s context and history.