vitas/evidra-lock
MCP Kill-switch for AI agents. Validates infrastructure operations before execution. Fail-closed. Evidence-backed.
Evidra-Lock acts as a critical safety net when AI agents or CI pipelines interact with your production infrastructure. It takes structured commands intended for tools like kubectl or Terraform and checks them against predefined safety policies. If a command is dangerous, incomplete, or unknown, it's blocked, preventing unintended outages or security issues. Operations engineers and SREs use this to ensure AI agents only perform validated, safe actions on live systems.
Use this if you need an automatic, fail-closed safeguard to prevent AI agents or CI pipelines from executing unsafe or unauthorized infrastructure changes.
Not ideal if you need to analyze raw natural language commands from an AI, as it requires structured tool invocations.
Stars
7
Forks
—
Language
Go
License
Apache-2.0
Category
Last pushed
Mar 10, 2026
Commits (30d)
0
Get this data via API
curl "https://pt-edge.onrender.com/api/v1/quality/mcp/vitas/evidra-lock"
Open to everyone — 100 requests/day, no key needed. Get a free key for 1,000/day.
Higher-rated alternatives
AndrewAltimit/template-repo
Agent orchestration & security template featuring MCP tool building, agent2agent workflows,...
knowledgepa3/gia-mcp-server
MCP proxy for GIA Governance — connects Claude Desktop and Claude Code to the hosted GIA...
Chimera-Protocol/csl-core
Deterministic safety layer for AI agents. Z3-verified policy enforcement.
portofcontext/pctx
pctx is the execution layer for agentic tool calls. It auto-converts agent tools and MCP servers...
agentralabs/agentic-contract
Policy engine for AI agents — enforceable rules, risk limits, approval gates, obligation...