humanforai

humanforai

MCP server for Human For AI — a human endpoint for AI agents. Let your AI agent hire a real human operator for tasks that need physical presence, human perception, or human judgment:

• Real-world verification — confirm a place, product, price, or claim exists, with photo/text evidence
• Product or app testing — a real human installs, uses, and reports
• Human judgment and feedback — tone, clarity, trustworthiness, design quality
• AI output review — human review before your output reaches production
• Data collection — gathering/labeling that needs human perception or local access
• Local physical-world tasks — visit, photograph, check, measure, observe
• …and more — call get_human_services for the full catalog

All services are free during the proof-of-concept pilot. No API key, no payment. Every task is reviewed by the human operator before acceptance; illegal, harmful, deceptive, unsafe, or privacy-invasive tasks are rejected. First response within 12 hours on working days (Sun–Thu).

Tools

Tool	Purpose
get_human_services	Service catalog, accepted task types, limits
submit_human_task	Submit a task; returns a task_id for status polling
check_task_status	Poll a submitted task by task_id
message_human_operator	Ask questions / scope work before submitting

Setup

Claude Code

claude mcp add human-for-ai -- npx -y humanforai

Claude Desktop / other MCP clients (stdio)

{
  "mcpServers": {
    "human-for-ai": {
      "command": "npx",
      "args": ["-y", "humanforai"]
    }
  }
}

Clients with native remote MCP support

You don't need this package — connect directly to the hosted endpoint (streamable HTTP, no auth):

https://humanforai.dev/mcp

E.g. Claude (Settings → Connectors → Add custom connector), or:

claude mcp add --transport http human-for-ai https://humanforai.dev/mcp

How it works

This package is a thin stdio proxy to the hosted Human For AI MCP endpoint. Tool definitions come from the live server, so they are always current — the package itself has no logic to go stale.

Previously published as human-api, which still works but is deprecated in favor of this package.

Typical flow

• get_human_services — see what the operator can do
• submit_human_task — be specific and self-contained; include contact_email to receive the deliverable
• check_task_status — poll with the returned task_id

Links

• Website: https://humanforai.dev
• API docs: https://humanforai.dev/api
• Machine-readable manifest: https://humanforai.dev/.well-known/agent.json
• OpenAPI spec: https://humanforai.dev/openapi.json
• Trust & safety: https://humanforai.dev/trust

License

MIT