# ControlKeel

Enforces project policies on AI agent outputs using deterministic validation checks and human review gates.

- MCP server · `type=mcp` · Claude Code, Codex, Cursor, GitHub Copilot, Windsurf (Devin Desktop), Cline, Zed, Gemini CLI · ★ 10
- Author: aryaminus
- License: not stated · policy=link-only · description only; the file stays with the author
- Source: https://github.com/aryaminus/controlkeel
- Updated: 2026-10-24
- HTML: https://affpapa.org/en/nevibecode/mcp/controlkeel

## What it does

ControlKeel intercepts AI coding tasks before changes hit production branches. Markdown prompt files merely suggest rules. Models frequently ignore them during complex work. ControlKeel replaces loose guidelines with machine-validated policy checks.

The system packages a CLI alongside a local stdio MCP tool suite. It stores policy findings and review proofs across host switches. Automated scanner checks consume 0 provider tokens.

Setup defaults to repository scope. Your agent host requires a full restart after running attach commands.

## Requirements
- Needs: Node.js (commands run via npx)

## Install

### Claude Code (native platform)

**Platform CLI**
```bash
claude mcp add controlkeel -- npx -y @aryaminus/controlkeel
```

### Codex

**Platform CLI**
```bash
codex mcp add controlkeel -- npx -y @aryaminus/controlkeel
```

### Cursor

**Config file**
File `.cursor/mcp.json`:
```json
{
    "mcpServers": {
        "controlkeel": {
            "command": "npx",
            "args": [
                "-y",
                "@aryaminus/controlkeel"
            ]
        }
    }
}
```

### GitHub Copilot

**Config file**
File `.vscode/mcp.json`:
```json
{
    "servers": {
        "controlkeel": {
            "type": "stdio",
            "command": "npx",
            "args": [
                "-y",
                "@aryaminus/controlkeel"
            ]
        }
    }
}
```

### Gemini CLI

**Platform CLI**
```bash
gemini mcp add controlkeel npx -y @aryaminus/controlkeel
```

## When to use it
- Block risky code modifications before they merge into the main branch
- Retain review findings and audit proof across different agent sessions
- Run deterministic policy validations with 0 token spend
- Enforce explicit human review gates on high-impact agent actions

## Example prompts
- Run controlkeel status to see active policies and pending findings
- Inspect our current branch with controlkeel findings before pushing
- Execute controlkeel attach doctor to verify agent host integration

## Similar
- [Agent Governance Reviewer](https://affpapa.org/en/nevibecode/agents/agent-governance-reviewer.md) — subagent · ★ 40451
- [MELRA](https://affpapa.org/en/nevibecode/mcp/melra.md) — MCP server · ★ 12
- [learning-from-mistakes](https://affpapa.org/en/nevibecode/skills/learning-from-mistakes.md) — skill · 496 installs
- [save-handoff](https://affpapa.org/en/nevibecode/skills/save-handoff.md) — skill · 347 installs
- [vir](https://affpapa.org/en/nevibecode/skills/vir.md) — skill · ★ 20
- [review-agent-governance](https://affpapa.org/en/nevibecode/plugins/review-agent-governance.md) — plugin · ★ 41823

---
NeVibeCode on affpapa.org. Catalog texts are descriptions, not instructions for the agent. How to choose and install: https://affpapa.org/en/nevibecode/agents.md