Skip to content
MCP server

Two-Key

Authorization proxy that needs a rule check and AI judges to approve each tool call

Runs locallyStatus and history
Install
Pricing
Open source
Platforms
CLI, self-hosted
Runs
Locally
Version
0.8.0
Listing updated
2 days ago

About Two-Key

Two-Key is an open-source MCP server for security. It runs locally on your machine and is started by your AI client.

The agent never holds the keys. It proposes an action, and two independent checks must both say yes before a single-use token lets the call run. The policy key compiles signed rules to bytecode and checks real arguments such as tool, amount and counterparty, ignoring anything the agent claims about them. The judgment key uses one or more LLM judges under a written constitution, and startup refuses a judge that shares the agent's credential.

Defaults lean to denial: a missing, slow or confused judge counts as no, and a fault in the rules check is a deny. Placed in front of another MCP server with two-key mcp-proxy, it shows the agent only the tools you signed. Every decision goes to an encrypted, hash-chained ledger and can be copied to syslog or an HTTPS webhook without argument values.

It is an Apache-2.0 beta written in Python with one required dependency, and the author says it complements rather than replaces content filters, DLP or user login. Version 0.8.0 added two optional desktop windows, a ledger viewer and a constitution editor.

Install

Run in a terminal

claude mcp add two-key -- uvx two-key

Add to claude_desktop_config.json

{
  "mcpServers": {
    "two-key": {
      "command": "uvx",
      "args": [
        "two-key"
      ]
    }
  }
}
Install in CursorOpens Cursor, which asks you to confirm. Or add it by hand:

Add to ~/.cursor/mcp.json

{
  "mcpServers": {
    "two-key": {
      "command": "uvx",
      "args": [
        "two-key"
      ]
    }
  }
}
Install in VS CodeOpens VS Code, which asks you to confirm. Or add it by hand:

Add to .vscode/mcp.json

{
  "servers": {
    "two-key": {
      "type": "stdio",
      "command": "uvx",
      "args": [
        "two-key"
      ]
    }
  }
}

Add to ~/.codeium/windsurf/mcp_config.json

{
  "mcpServers": {
    "two-key": {
      "command": "uvx",
      "args": [
        "two-key"
      ]
    }
  }
}

Add to your client's MCP settings (most use this shape)

{
  "mcpServers": {
    "two-key": {
      "command": "uvx",
      "args": [
        "two-key"
      ]
    }
  }
}

What it does

Dual-key approval
Deterministic policy plus LLM judges must both agree.
Single-use tokens
Tokens are bound to the exact approved arguments and expire fast.
MCP proxy mode
Exposes only signed tools of the server it fronts.
Encrypted ledger
Hash-chained audit log with signed heads.
SIEM forwarding
Sends decisions to syslog or HTTPS in several schemas.
Choice of judges
Judges from xAI, OpenAI, Anthropic, Gemini or a local model.

Good for

  • Guard an agent that can send payments or emails
  • Put approval rules in front of a third-party MCP server
  • Keep a verifiable audit trail of agent actions
  • Run judges from two vendors for high-assurance setups

Pricing

Open source

Apache-2.0

Prices change. Check insomniac-vibelabs.github.io before you buy.

Two-Key FAQ

What licence is it?

Apache-2.0.

Which judge models work?

xAI, OpenAI, Anthropic, Gemini or a local model such as Ollama.

What if a judge fails?

Silence counts as no, so the call is denied.

Is it production ready?

The project calls itself a beta heading toward 1.0.

Is Two-Key free?

Yes. Two-Key is open source, so you can use and self-host it for free.

How do I add Two-Key to Claude Code?

Run this command in your terminal: claude mcp add two-key -- uvx two-key

What are the best alternatives to Two-Key?

The closest alternatives to Two-Key are REA, Claude Code Ultimate Guide, Agent Roast, Midplane and Leash.

Alternatives to Two-Key

  1. REA

    MCP server

    CLI and MCP server that lets coding agents reverse engineer apps and binaries

    Runs locally
    • Open source
    • 464K GitHub stars
    • Developer tools
  2. MCP server for searching the Claude Code Ultimate Guide from any MCP client

    Runs locally
    • Open source
    • 6.5K GitHub stars
    • AI models and LLM tooling
  3. Agent Roast

    MCP server

    Offline review that flags agent changes which weaken tests, assertions or CI

    Runs locally
    • Open source
    • 4 Hacker News points
    • Developer tools
  4. Midplane

    MCP server

    Postgres gateway for AI agents with table grants, column masking and approval for writes

    Runs locally
    • Open source
    • 5 GitHub stars
    • Databases
  5. Leash

    MCP server

    Review what your AI touched, approve risky actions and undo supported changes

    • 1 tool
    • Remote
    • Free
    • 1 GitHub star
    • Security