"author:COSTRINITY" matching MCP servers:

A containment layer for AI agents that go rogue. Guard mode wraps stdio MCP servers: calls to the tools you mark as risky wait for a person; if nobody decides in time, the call is blocked. No API key can decide a hold, including the agent’s own. Claimed accounts get Ed25519-signed session records you verify offline. Covers only the servers you wrap; detection is heuristic. Free tier, MIT client.