REA
Verified · 33 days agoReverse engineer anything from your terminal or agent with one CLI and MCP server.
Fail-closed Cedar policy gate + Ed25519 signed receipts for agent tool calls. Denies on any error.
claude mcp add protect-mcp -- npx -y [email protected]
Policy gateway that wraps any MCP server and evaluates every tool call against a Cedar policy (AWS's IAM language), denying on any error, missing engine, or evaluation failure — it even refuses to start in enforce mode unless a startup self-test proves a known-forbidden action is denied. Each decision gets an Ed25519-signed, offline-verifiable receipt, and the workflow is sensible: shadow-mode observation, a localhost dashboard with an approval queue, auto-drafted policies, then enforcement. MIT, TypeScript, npx-runnable and local-first, but it is a three-person shop with 8 stars, the '3 IETF Internet-Drafts' are self-authored submissions, and the ~1,850 npm downloads/week look mirror-inflated given 30 releases since March; the paid ScopeBlind service adds hosted digest anchoring, though the free local gate stands alone. For teams putting agents in front of consequential tools who want deny-by-default enforcement with a cryptographic audit trail.
Reverse engineer anything from your terminal or agent with one CLI and MCP server.
Agent teams, swarms, memory and hooks for Claude Code and Codex, served over MCP.
AI orchestration with hive-mind swarms, neural networks, and 87 MCP tools for enterprise dev.
Draw and visually collaborate with your agents on tldraw's canvas.
Codebase knowledge graph for AI agents — 162 languages, sub-ms queries, 99% fewer tokens.
Keyword research, site audits, rank tracking and AI visibility for your AI agent.