{
  "markdown": "# Humwork MCP Plugin\n\nConnect AI agents with verified domain experts in real-time. When your AI agent needs human expertise — for debugging, architecture decisions, or unfamiliar domains — Humwork matches it with a specialist who diagnoses the problem and provides solutions, all within the agent's context.\n\n## Installation\n\n### Claude Code\n\n#### Plugin\n\nInside Claude Code, run:\n\n```bash\n/plugin marketplace add humworkai/humwork-mcp\n```\n\nThen install the plugin:\n\n```bash\n/plugin install humwork@humwork-mcp\n```\n\nFinally, run `/mcp`, select **humwork**, click **Re-authenticate**, and sign in at [humwork.ai](https://humwork.ai). Auth lasts 30 days.\n\n#### MCP\n\n```bash\nclaude mcp add --transport http humwork https://api.humwork.ai/api/v1\n```\n\nOr add manually to your project's `.mcp.json`:\n\n```json\n{\n  \"mcpServers\": {\n    \"humwork\": {\n      \"type\": \"http\",\n      \"url\": \"https://api.humwork.ai/api/v1\"\n    }\n  }\n}\n```\n\n**Authentication:** Run `/mcp`, select **humwork**, click **Re-authenticate**, and sign in at [humwork.ai](https://humwork.ai). Auth lasts 30 days.\n\n---\n\n### Claude.ai (Web)\n\n1. Go to [claude.ai/customize/connectors](https://claude.ai/customize/connectors)\n2. Click the **+** button to add a custom connector\n3. Fill in:\n   - **Name:** `Humwork`\n   - **Remote MCP server URL:** `https://api.humwork.ai/api/v1`\n4. Click **Add** — a browser window will open to authenticate with your Humwork account\n5. Once connected, set tool permissions to **Always allow** or **Needs approval** as preferred\n\nAll 6 tools will be available: `consult_expert`, `send_chat_message`, `get_chat_messages`, `extend_session`, `close_chat`, and `rate_chat`.\n\n---\n\n### Cursor\n\nAdd to your project's `.cursor/mcp.json` (or global `~/.cursor/mcp.json`):\n\n```json\n{\n  \"mcpServers\": {\n    \"humwork\": {\n      \"url\": \"https://api.humwork.ai/api/v1\",\n      \"headers\": {\n        \"X-API-Key\": \"YOUR_API_KEY\"\n      }\n    }\n  }\n}\n```\n\nOr: **Settings** (Cmd+Shift+J) → **MCP** → **Add new MCP Server** → paste the config above.\n\nGet your API key from your [Humwork dashboard](https://humwork.ai/dashboard).\n\n---\n\n### OpenAI Codex CLI\n\nRequires **Node.js 18+** (the bridge runs via `npx`). Check with `node --version`, install with `brew install node` if needed.\n\nAdd to `~/.codex/config.toml` (or `.codex/config.toml` in your project):\n\n```toml\n[mcp_servers.humwork]\ncommand = \"npx\"\nargs = [\n  \"-y\",\n  \"mcp-remote@latest\",\n  \"https://api.humwork.ai/api/v1\",\n  \"--header\",\n  \"X-API-Key:${HUMWORK_API_KEY}\"\n]\n\n[mcp_servers.humwork.env]\nHUMWORK_API_KEY = \"YOUR_API_KEY\"\n```\n\nGet your API key from your [Humwork dashboard](https://humwork.ai/dashboard).\n\n---\n\n### ChatGPT\n\nRequires ChatGPT **Plus, Pro, Business, or Enterprise** — custom apps are not available on the free tier.\n\n1. Go to **Settings → Apps → Advanced** and toggle **Developer mode** on. Accept the \"Elevated Risk\" warning.\n2. Back on the Apps page, click **Create app** (top right).\n3. Fill in:\n   - **Name:** `Humwork`\n   - **MCP Server URL:** `https://api.humwork.ai/api/v1/gpt`\n   - **Authentication:** `OAuth`\n   - Check **\"I understand and want to continue\"** and click **Create**.\n4. A browser window will open — sign in with your Humwork account to authorize.\n\n---\n\n### Windsurf\n\nAdd the MCP server in Windsurf settings:\n\n```json\n{\n  \"mcpServers\": {\n    \"humwork\": {\n      \"url\": \"https://api.humwork.ai/api/v1\",\n      \"headers\": {\n        \"X-API-Key\": \"YOUR_API_KEY\"\n      }\n    }\n  }\n}\n```\n\nOr: **Settings** → **MCP / Plugins** → **Add MCP Server** → paste the config above.\n\nGet your API key from your [Humwork dashboard](https://humwork.ai/dashboard).\n\n---\n\n### Lovable\n\n1. Open **Settings** → **Connectors** → **Personal connectors**\n2. Click **New MCP server**\n3. Name: `humwork`, Server URL: `https://api.humwork.ai/api/v1`\n4. Authenticate with your Humwork account when prompted\n\n---\n\n### OpenClaw\n\nRequires **OpenClaw 2026.4+** and **Node.js 22+**. Install with `npm install -g openclaw@latest` if needed.\n\n1. Add Humwork as an MCP server:\n\n```bash\nopenclaw mcp set humwork '{\"url\":\"https://api.humwork.ai/api/v1\",\"transport\":\"streamable-http\",\"headers\":{\"X-API-Key\":\"YOUR_API_KEY\"}}'\n```\n\n2. Restart the gateway to pick up the config:\n\n```bash\nopenclaw gateway stop && openclaw gateway\n```\n\n3. Verify and open the dashboard:\n\n```bash\nopenclaw mcp list\nopenclaw dashboard\n```\n\nGet your API key from your [Humwork dashboard](https://humwork.ai/dashboard).\n\n> **Note:** OpenClaw requires a model provider API key configured separately. Run `openclaw models auth paste-token --provider anthropic` to add your Anthropic key.\n\n## Available Tools\n\n| Tool | Description |\n|------|-------------|\n| `consult_expert` | Start a consultation with a matched expert. Describe the problem, what you've tried, and relevant context. |\n| `send_chat_message` | Send a follow-up message in an active expert chat session. |\n| `get_chat_messages` | Retrieve messages from an expert chat session. |\n| `close_chat` | Close a chat session when the problem is resolved. Optionally include a rating (1-5). |\n| `rate_chat` | Rate an expert's helpfulness (1-5) after a chat session closes. Must be submitted within 15 minutes. |\n\n## How It Works\n\n1. Your AI agent identifies a need for expert guidance (repeated failures, architectural decisions, unfamiliar domains)\n2. It calls `consult_expert` with problem details, attempted solutions, and relevant context\n3. Humwork matches the request with a verified expert based on domain expertise\n4. The expert and agent chat in real-time until the problem is resolved\n5. The agent calls `close_chat` to end the session (with an optional rating)\n6. If no rating was provided at close, the agent calls `rate_chat` within 15 minutes\n\n## What's Included\n\n- **Plugin manifest** (`.claude-plugin/plugin.json`) — name, version, description\n- **MCP server config** (`.mcp.json`) — server URL\n- **Platform configs** (`configs/`) — ready-to-use configs for Cursor, Codex, Windsurf, and Lovable\n- **Hooks** (`hooks/`) — behavioral hooks that teach agents when to escalate\n- **Skills** (`skills/`) — guidance that helps agents escalate effectively\n- **Commands** (`commands/`) — setup instructions for authentication\n\n## Links\n\n- [Humwork](https://humwork.ai) — Homepage\n- [Documentation](https://humwork.ai/docs) — Full docs\n\n## License\n\nMIT\n",
  "bytes": 6312,
  "sha": "5368be2902450f0824e5938c46d0c3a0e0285158ca2115c17be165b6203e0cf4",
  "repo_slug": "humworkai/humwork-mcp",
  "fonte": "repo",
  "truncated": false,
  "api": "https://agentalog.com/api/listings/plg_humworkai_humwork_mcp_humwork_8ad56960/readme"
}