AI agent governance MCP servers

A server here enforces policy on, approves, audits or restricts what AI agents are allowed to do.

Not here: servers that are merely used by agents.

1,891servers
748 remote1,143 localTagged by Odel's AI, which can make mistakes

Best fit

1,145 made for this, ranked by starsRemoteLocal
  1. RLHF feedback loop @igorganapolsky/rlhf-feedback-loopRLHF feedback loop for AI agents. Capture signals, promote memories, block mistakes, export DPO.8% chance this is malicious. Assessed by AI, which can make mistakes.Runs on your machine, installed from npm.28 GitHub stars.
  2. Humane Proxy @vishisht16/humane-proxyAI safety middleware — detects self-harm and criminal intent in LLM prompts.5% chance this is malicious. Assessed by AI, which can make mistakes.Runs on your machine, installed from PyPI.28 GitHub stars.
  3. Roast My Design System @gregkozakiewicz/roast-my-design-systemAudit your Design System and serve your Agent the rules that keep AI-written UI on-system.3% chance this is malicious. Assessed by AI, which can make mistakes.Runs on your machine, installed from npm.27 GitHub stars.
  4. RepoRelay @lukie-81/reporelayBounded local MCP access to one approved repository for AI read/search and structured handoffs.5% chance this is malicious. Assessed by AI, which can make mistakes.Runs on your machine, installed from npm.27 GitHub stars.
  5. KittyClaw @ekioo/kittyclawDrive KittyClaw, a self-hosted agent-fleet kanban, from any MCP client (embedded /mcp endpoint).4% chance this is malicious. Assessed by AI, which can make mistakes.Runs on your machine, installed from a package registry.27 GitHub stars.
  6. Rigour @rigour-labs/rigourLets coding agents review their change, find an existing helper, and recall the team's lessons.5% chance this is malicious. Assessed by AI, which can make mistakes.Runs on your machine, installed from npm.27 GitHub stars.
  7. CTRLRun Operator @ctrlrun/ctrlrun-mcp-operatorAnswer CTRLRun approvals from an MCP client: list what is held, inspect it, grant or deny.3% chance this is malicious. Assessed by AI, which can make mistakes.Runs on your machine, installed from PyPI.26 GitHub stars.
  8. Audrey @evilander/audreyLocal-first biological memory for coding agents: recall, consolidation, and pre-action Guard checks.5% chance this is malicious. Assessed by AI, which can make mistakes.Runs on your machine, installed from npm.26 GitHub stars.
  9. Speclock @sgroy10/speclockAI Constraint Engine — enforces CLAUDE.md and .cursorrules as laws. 51 MCP tools.6% chance this is malicious. Assessed by AI, which can make mistakes.Runs on your machine, installed from npm.25 GitHub stars.
  10. Delego @delego-dev/delegoIntent-bound action authorization for AI agents: policy, human approval, and a signed audit trail.3% chance this is malicious. Assessed by AI, which can make mistakes.Runs on your machine, installed from PyPI.25 GitHub stars.
  11. McpMux @mcpmux-io/mcp-muxMcpMux — configure MCP servers once, connect every AI client through one local gateway.5% chance this is malicious. Assessed by AI, which can make mistakes.Runs on your machine, installed from a package registry.24 GitHub stars.
  12. Mneme Decision MCP @mnemehq/mnemeLocal MCP server for architectural decisions, traceability, and non-authoritative proposals.3% chance this is malicious. Assessed by AI, which can make mistakes.Runs on your machine, installed from PyPI.24 GitHub stars.
  13. Fable Mode @rex-codebase/fable-modeMechanical write gates, evidence receipts, and adversarial review for AI coding agents.5% chance this is malicious. Assessed by AI, which can make mistakes.Runs on your machine, installed from PyPI.24 GitHub stars.
  14. Delimit — Keep the state. Change the model. @delimit-ai/delimit-mcp-serverKeep the state. Change the model. Shared context and a multi-model panel across coding assistants.8% chance this is malicious. Assessed by AI, which can make mistakes.Runs on your machine, installed from npm.24 GitHub stars.
  15. McpMux @mcpmux/mcp-muxMcpMux — configure MCP servers once, connect every AI client through one local gateway.6% chance this is malicious. Assessed by AI, which can make mistakes.Runs on your machine, installed from a package registry.24 GitHub stars.
  16. World Model @saravananjaichandar/world-model-mcpTemporal knowledge graph for codebases with constraint enforcement at the edit boundary.5% chance this is malicious. Assessed by AI, which can make mistakes.Runs on your machine, installed from PyPI.23 GitHub stars.
  17. Strategic Agent Reasoning MCP @putervision/agent-reasoning-mcpStrategic BDI reasoning engine for autonomous AI agents with spatial utility and replanning.5% chance this is malicious. Assessed by AI, which can make mistakes.Runs on your machine, installed from npm.21 GitHub stars.
  18. Claude Octopus @xiaolai/claude-octopusSpawn multiple specialized Claude Code agents as MCP servers, each independently configured.5% chance this is malicious. Assessed by AI, which can make mistakes.Runs on your machine, installed from npm.21 GitHub stars.
  19. Mastyf Ai @mastyf-ai/mastyf-aiRuntime proxy for MCP security, cost governance & audit7% chance this is malicious. Assessed by AI, which can make mistakes.Runs on your machine, installed from npm.21 GitHub stars.
  20. Codex Octopus @xiaolai/codex-octopusSpawn multiple specialized Codex agents as MCP servers, each independently configured.6% chance this is malicious. Assessed by AI, which can make mistakes.Runs on your machine, installed from npm.20 GitHub stars.
  21. Context Foundry @context-foundry/context-foundryRecursive Claude spawning for autonomous project builds with self-healing test loops8% chance this is malicious. Assessed by AI, which can make mistakes.Runs on your machine, installed from PyPI.20 GitHub stars.
  22. DOS — the trust substrate for agent fleets @anthony-chaudhary/dos-kernelVerify what agents actually shipped, arbitrate file collisions, refuse with structured reasons.4% chance this is malicious. Assessed by AI, which can make mistakes.Runs on your machine, installed from PyPI.20 GitHub stars.
  23. Ekbasis @openinterpretability/ekbasisForesee what git commands and multi-step changes will do, as calibrated probabilities3% chance this is malicious. Assessed by AI, which can make mistakes.Runs on your machine, installed from PyPI.19 GitHub stars.
  24. Sdd @juanklagos/sdd-mcpSpec-Driven Development workflow: specs, plans, gates, status and logbook tools. Bilingual EN/ES.4% chance this is malicious. Assessed by AI, which can make mistakes.Runs on your machine, installed from npm.19 GitHub stars.