AI导哈

RentAHuman 任务编排器

一个MCP服务器，让AI智能体将复杂任务分解为并行的RentAHuman.ai赏金任务，并以单一编排统一管理，适用于需要人类执行子任务的AI工作流。

  • by brandonore
  • TypeScript
  • 本地部署
  • 更新于
在 GitHub 查看 brandonore/rentahuman-orchestrator

使用说明

rentahuman-orchestrator

An MCP server that lets AI agents break complex tasks into parallel RentAHuman.ai bounties and manage them as a single orchestration.

AI agent does the thinking. This server does the plumbing.

Install

npx rentahuman-orchestrator

Add to Claude Code

Mock mode — searches real humans, bounties are simulated:

{
  "mcpServers": {
    "rentahuman-orchestrator": {
      "command": "npx",
      "args": ["rentahuman-orchestrator"],
      "env": {
        "MOCK_MODE": "true"
      }
    }
  }
}

Real mode — everything is live, requires an API key:

{
  "mcpServers": {
    "rentahuman-orchestrator": {
      "command": "npx",
      "args": ["rentahuman-orchestrator"],
      "env": {
        "RENTAHUMAN_API_KEY": "rah_your_key_here"
      }
    }
  }
}

Modes

Search humans Create bounties Accept applicants
Mock (MOCK_MODE=true) Real API Simulated Simulated
Real (with API key) Real API Real API Real API

Mock mode is safe for testing — you'll see real people from the platform but nothing gets posted or charged.

Tools

Tool What it does
smart_match Search and rank real humans by skill, location, and rate
orchestrate_task Create multiple bounties from a list of subtasks
get_orchestration Check status of all bounties in an orchestration
cancel_orchestration Cancel all active bounties
review_applications View applicants and accept the best fits

Example

"I'm organizing a food pop-up event in Tokyo. I need a photographer, videographer, social media person, and event staff."

The agent searches all 4 roles in parallel using smart_match and returns real people from the platform:

Results from real API data:

Role Best Match Rate Location Notes
Photographer Japanese_food_boy $20/hr Edogawa, Tokyo Food photography + Japanese cuisine
Videographer 成田U輔 $50/hr Setagaya, Tokyo Photography + videography
Social Media Denjin $60/hr Tokyo Social media + events + production
Event Staff Denjin $60/hr Tokyo Multi-role — covers events + photography

Then creates bounties for all roles in one call with orchestrate_task, with a cost breakdown:

Role Bounty ID Budget Hours
Photographer mock_bounty_1 $100 4 hrs
Videographer mock_bounty_2 $100 4 hrs
Social Media mock_bounty_3 $100 4 hrs
Event Staff mock_bounty_4 $100 4 hrs
Total $400 16 hrs

Check on applicants with review_applications, accept the best fits, and you're set.

Get an API Key

  1. Sign up at rentahuman.ai/signup
  2. Get verified ($9.99/mo)
  3. Generate a key at Dashboard → API Keys

Your key starts with rah_. Set it as an environment variable rather than hardcoding it in the MCP config:

export RENTAHUMAN_API_KEY=rah_your_key_here

The MCP config (~/.claude.json) is a local file and not committed to git, but using an env var or .env file is still the safer habit.

Development

git clone https://github.com/brandonore/rentahuman-orchestrator.git
cd rentahuman-orchestrator
bun install
bun run build

Test with MCP Inspector:

npx @modelcontextprotocol/inspector node build/index.js

Architecture

AI Agent (Claude)           →  rentahuman-orchestrator (MCP)  →  RentAHuman.ai API
Decomposes task                Batch creates bounties            Handles individual
into subtasks                  + tracks state + caches           bookings & searches

License

MIT

相关 MCP