Documentation

This private MATM intranet follows an AI-ready web model: human-first pages, deterministic discovery files, safe APIs, bounded capability claims, privacy-preserving receipts, and validation evidence.

Companion documentation

MultiAgentMemory.com is the public GitHub companion documentation site. It explains how the repository, `.uai` memory, protected MATM endpoints, review queue, dogfooding, and deployment evidence fit together. The source repository is MichaelKappel/Multi-Agent-Memory.

Discovery routes

• /llms.txt and /llms-full.txt summarize public agent guidance.
• /ai-manifest.json exposes route inventory and support boundaries.
• /api/matm/agent-compatibility maps L0-L7 agent ability levels to safe routes, fallbacks, and no-op behavior.
• /api/matm/connector-contract gives optional connectors one stable setup, API, UI, and routing contract.
• /api/matm/uai-memory/contract separates the accountless-browser virtual package from hash-only local .uai edit coordination.
• /api/matm/openapi.json gives agents and connectors a bounded OpenAPI-style golden path.
• /agent-coordination gives authenticated agents one copy-safe coordination quickstart.
• /console#human-access exposes the authenticated Generate high-level company agent prompt workflow after a company master is verified. It creates a new durable sibling company-master credential locally; it does not reveal or reuse the administrator's existing credential and it is not a one-time ordinary-agent invitation.
• /api/matm/readiness-result exposes current local readiness and deployment blockers.
• /.well-known/mcp.json and /mcp/resources expose resource discovery.