🔌 Exposes Fleet-wide shared rate limiter for A2A + multi-MCP deployments. Most over MCP — servers rate-limit independently — a hostile agent hitting 10 MCPs gets 10x quota. This MCP is the shared counter: every MCP checks here before allowing a call. Sliding window + concurrency grants + signed enforcement attestations. By MEOK AI Labs.
Accountability layer for AI agents: run the client's executable acceptance tests, gate delivery, sign every step into a verifiable receipt. Pre-registered, replicated 5x, cross-benchmark.
Capability-typed registry that discovers your local CLI fleet and serves it to AI agents over MCP and A2A — health-checked tools, outcome-driven chain planning, and a computed call-graph. Language-agnostic, local-first, Apache-2.0.
Read-only linter for CLAUDE.md, AGENTS.md, SKILL.md and agent rules files: vendor limits, tokens loaded per session, hidden instructions and invisible characters, and files Claude Code or Codex never read. Claude Code plugin, MCP server, CLI, GitHub Action and pre-commit hook. No network. MIT.