Metadata-Version: 2.4
Name: ai2human-mcp
Version: 0.1.1
Summary: AI2Human MCP server — Claude can dispatch human-execution tasks with proof verification.
Author: AI2Human
License: MIT
Keywords: agent,ai2human,claude,human-in-the-loop,mcp
Classifier: Development Status :: 3 - Alpha
Classifier: Intended Audience :: Developers
Classifier: License :: OSI Approved :: MIT License
Classifier: Programming Language :: Python :: 3
Classifier: Programming Language :: Python :: 3.10
Classifier: Programming Language :: Python :: 3.11
Classifier: Programming Language :: Python :: 3.12
Requires-Python: >=3.10
Requires-Dist: fastmcp<4,>=3.0
Requires-Dist: httpx<1,>=0.27
Description-Content-Type: text/markdown

# AI2Human MCP Server

Claude can dispatch human-execution tasks with proof verification.

When your Claude agent hits a step it can't complete alone — local verification, physical checks, identity actions, compliance reviews — AI2Human dispatches it to a human operator with cryptographic proof verification.

## Quickstart

```bash
pip install ai2human-mcp
```

Set your API key:

```bash
export AI2HUMAN_API_KEY="your-key"
```

Run:

```bash
ai2human-mcp
```

## Claude Desktop Config

```json
{
  "mcpServers": {
    "ai2human": {
      "command": "uvx",
      "args": ["ai2human-mcp"],
      "env": {
        "AI2HUMAN_API_KEY": "YOUR_KEY"
      }
    }
  }
}
```

## Tools

| Tool | Description |
|------|-------------|
| `ai2human_list_categories` | List supported task categories and proof types |
| `ai2human_create_task` | Dispatch a human-execution task |
| `ai2human_check_task` | Check task status, verification, and settlement |
| `ai2human_get_proof` | Retrieve evidence bundle and settlement receipt |

## Environment

| Variable | Required | Default |
|----------|----------|---------|
| `AI2HUMAN_API_KEY` | Yes | — |
| `AI2HUMAN_AGENT_KEY` | No | Fallback for API key |
| `AI2HUMAN_BASE_URL` | No | `https://ai2human.io` |

<!-- mcp-name: io.github.ai2humanagent/ai2human-mcp -->
