Marrow AI
Enforce policy controls, pre-action approvals, and reusable audit proof across autonomous multi-agent workflows.
Marrow AI acts as a runtime governance and control layer for autonomous AI agents across MCP-compatible clients such as Claude Code, Cursor, and Codex. By requiring verified approvals before high-stakes operations—such as production deployments, package publishing, or credential access—it prevents unauthorized execution while recording verifiable evidence and operational lessons that downstream agents can reuse.
主な機能
01Implements pending write reconciliation with idempotent retries and request hashing
02Maintains durable audit trails and reusable outcome proof across multi-agent fleets
03Offers flexible tool profiles to customize the exposed MCP tool surface
047 GitHub stars
05Enforces pre-action policy verification and human approvals before critical actions execute
06Provides a keyless seatbelt guard to block secret leaks, force-pushes, and destructive deletions
ユースケース
01Establishing accountability and proof-backed outcome verification in shared agent workflows
02Safeguarding production deployments and package publishing from unintended autonomous agent actions
03Preventing sensitive credentials and environment files from being read or exposed by coding assistants