Control and audit AI agent access to MCP servers using CEL-based policies, role-based access control, and comprehensive logging. For teams deploying Claude or other AI agents that need security governance and compliance tracking.
io.github.Sentinel-Gate/sentinelgate
Local install
STDIO
No auth required
How models use it and what it is built for.
Control and audit AI agent access to MCP servers using CEL-based policies, role-based access control, and comprehensive logging. For teams deploying Claude or other AI agents that need security governance and compliance tracking.
Local install — runs as a subprocess.
Where to find authoritative docs and source for SentinelGate.
Paste any of these into Agent Studio after connecting SentinelGate.
Common questions about connecting and running SentinelGate.
What is SentinelGate and why would I use it?
SentinelGate is an open-source MCP proxy that sits between AI agents and MCP servers, enforcing access control policies. Use it when you need to restrict what tools or data your Claude instance can access, audit agent actions, or implement role-based permissions.
How do I define access policies?
SentinelGate uses CEL (Common Expression Language) for policy definition, allowing you to write expressive rules like 'deny if tool == database AND user_role != admin'. The install includes the v2.0 binary; see the GitHub repository for policy syntax examples.
Does SentinelGate log all agent activity?
Yes, SentinelGate provides audit logging of all access attempts, including denied requests. This is useful for compliance, debugging, and understanding what your agents are doing.
What authentication methods does it support?
SentinelGate implements role-based access control (RBAC) for permission management. Refer to the GitHub repository for details on how to configure roles and integrate with your identity provider.
Is SentinelGate production-ready?
SentinelGate is open-source and at v2.0, indicating active development. Review the GitHub repository and test thoroughly in your environment before deploying to production, especially for security-critical use cases.
MCP Playground runs 10,000+ hosted MCP servers — GitHub, Linear, Notion, Stripe, Sentry and more — across Claude, GPT, Gemini, DeepSeek and 30+ AI models. Compare model answers side-by-side, save agent presets, share runs. Zero install.
Open Agent Studio