elliot35

elliot35 / deterministic-agent-control-protocol

Public

Governance gateway for AI agents — bounded, auditable, session-aware control with MCP proxy, shell proxy & HTTP API. Works with Cursor, Claude Code, Codex, and any MCP-compatible agent.

www.npmjs.compackage@det-acpcore agent-control agent-governance ai-agent ai-governance ai-safety
131
5
100% credibility
Found Feb 09, 2026 at 12 stars 11x -- GitGems finds repos before they trend. Get early access to the next one.
Sign Up Free
AI Analysis
TypeScript
AI Summary

A safety gateway that enforces rules on AI agents' actions like file access and commands, making them auditable and reversible for tools like Cursor and Claude Code.

How It Works

1
🛡️ Discover Safe AI Control

You hear about a simple way to make your AI coding helper safe by setting rules on what it can touch.

2
📥 Add the Safety Layer

With one easy step, you bring the protector into your project folder.

3
✏️ Set Your Rules

You pick what your AI can do, like reading code files but never secrets or dangerous changes.

4
🔗 Link to Your AI Tool

You connect it to your favorite AI like Cursor or Claude, and it starts watching.

5
🤖 Watch AI Work Safely

Your AI suggests edits, runs safely allowed actions, and blocks risky ones instantly.

✅ Everything Logged & Reversible

You review the full story of what happened, undo anything if needed, and code confidently.

Sign up to see the full architecture

4 more

Sign Up Free

Star Growth

See how this repo grew from 12 to 131 stars Sign Up Free
Repurpose This Repo

Repurpose is a Pro feature

Generate ready-to-use prompts for X threads, LinkedIn posts, blog posts, YouTube scripts, and more -- with full repo context baked in.

Unlock Repurpose
AI-Generated Review

What is deterministic-agent-control-protocol?

This TypeScript-based AI governance gateway proxies AI agents like Cursor, Claude Code, or Codex, enforcing YAML policies on file reads/writes, shell commands, git ops, and HTTP requests to prevent secrets leaks or runaway sessions. It logs every action in a tamper-evident ledger, supports human approvals for risky moves, and offers rollback plans—all via MCP proxy, shell wrapper, or HTTP API. Think bounded, auditable agent control without rebuilding your workflow.

Why is it gaining traction?

One-command setup (`npx det-acp init cursor`) generates policies and configs for instant governance, unlike ad-hoc prompts or full sandboxes. Built-in production policies for coding, devops, and data analysis, plus session budgets and reversibility, make agents reliable for real projects. As an ai governance gateway on GitHub, it fills the gap for MCP-compatible agents needing explainable controls over loose api gateway governance tools.

Who should use this?

AI coding teams at startups deploying Cursor/Claude agents on shared repos, blocking .env reads or rm -rf bombs. DevOps folks automating deploys with bounded git commits and command runs. Indie devs prototyping agent-driven tools who want github governance framework without docker isolation headaches.

Verdict

Grab it for agent experiments—CLI shines, docs cover integrations thoroughly, tests hit core paths. 1.0% credibility and 13 stars signal early days; validate ledgers yourself before prod.

(198 words)

Sign up to read the full AI review Sign Up Free

Similar repos coming soon.