
HITL (Human in the Loop)
io.github.hitl-shPublished by GitHub account hitl-shv1.0.0Updated Nov 19, 2026
Let AI agents ask a real person for a decision on their phone and wait for the answer.
The publisher's own server: TAP
Overview
Lets an AI assistant run HITL workflows that ask a real person for a decision on their phone and wait for the answer.
- What you can do
- List, inspect and build workflows, including steps, connections, input fields and example input bodies
- Run, retry, test, activate and pause workflows, and list or wait on their runs
- Track runs with live-update URLs and published revisions
- See who reviews a workflow and invite reviewers by email
- Create, rotate, validate and join with workflow join codes
- Check the API key and handle agent signup status
- When to use it
- Use it when an agent needs a human judgment call mid-task, such as approving, choosing or confirming something, and should block until a person answers on the HITL mobile app. It suits agents that already have HITL workflows or want to build and manage them from the assistant.
- Requirements
- Remote endpoint at mcp.hitl.sh; OAuth sign-in is required for the main endpoint, while the simpler endpoint takes a HITL API key as a bearer token. A HITL account and the dedicated HITL mobile app are needed for reviewers to receive and complete tasks. A local stdio variant runs via npx with a HITL API key in the environment.
Installation
In SourceWeft
- Open HITL (Human in the Loop) in the dashboard and add it to a workspace.
- Connect the required credentials. They are stored privately and never shown on public pages.
- Enable the server for the chats that should use its tools.
Web executable via Streamable HTTP. Remote servers run from the web runtime once configured in a workspace.
Use with an agent
Paste this prompt into Claude Code, Codex, Cursor or another agent — or into SourceWeft chat. The agent reads this server's install guide, shows you its publisher, endpoint and sign-in, and adds it to your client once you agree. Keys never go into a command.
Read https://sourceweft.com/mcp/io.github.hitl-sh%2Fhitl/install.md and connect the MCP server it describes to my client. Before changing any config, show me its publisher, endpoint and sign-in method and wait for my OK. Never put an API key in a command or in a file I would commit; ask me where it should come from.Other MCP clients
Add this to your client's mcpServers config. The client completes OAuth sign-in on first connect.
{
"mcpServers": {
"hitl-human-in-the-loop": {
"type": "http",
"url": "https://mcp.hitl.sh/mcp"
}
}
}README
HITL.sh MCP Server
HITL.sh is a mobile-powered Human-in-the-Loop platform designed to give AI systems instant access to real human judgment. The entire system is built around our dedicated mobile app, which is required for the product to function and enables human reviewers to receive and complete tasks in real time.
🚀 Quick Start for Users
Choose the endpoint that matches your platform:
For ChatGPT Users (OAuth):
URL: https://mcp.hitl.sh/mcp
- Add MCP Connector in ChatGPT with URL above
- Sign up through the OAuth flow
- (Optional) Add your personal HITL API key at:
https://mcp.hitl.sh/setup-api-key - Start using HITL tools!
For OpenAI Agent Builder / Zapier / Make.com (Direct API Key):
URL: https://mcp.hitl.sh/mcp-simple
- Add MCP Server with URL above
- Authentication: Bearer Token
- Token: Your HITL API key (e.g.,
hitl_live_abc123...) - Start using HITL tools!
📚 Full OpenAI Agent Builder Guide
🛠️ For Developers
Features
- OAuth 2.1 Authentication - Secure OAuth flow with Auth0 integration
- Per-User API Keys - Users can configure their own HITL.sh API keys
- Workflow tools - Find, run, build and manage HITL workflows, and choose who reviews them
- ChatGPT Compatible - Works as a connector in ChatGPT
- Claude Desktop Support - Standalone stdio server for local use
- Production Ready - Deployed on Vercel with proper error handling
🛠️ Available Tools
HITL is workflows: an agent runs a workflow, the workflow asks your humans, and the run finishes with their answer.
Loops, requests, images, sessions, people, organizations, teams, webhooks and uploads are no longer tools (the API still serves them). The workflow routes are live on dev.api.hitl.sh and answer 404 on api.hitl.sh until they ship there; set HITL_API_BASE=https://dev.api.hitl.sh/v1 to use them now.
Both endpoints (/mcp and /mcp-simple) register these tools from @hitl-sh/core, the same package behind the stdio server @hitl-sh/mcp, so the remote and local servers expose identical tools.
📋 Prerequisites
- Node.js 18+ (download)
- HITL.sh API key (get one here)
- Auth0 account (sign up)
- Vercel account for deployment (sign up)
🔧 Environment Variables
Required for Production
Optional
🚀 Quick Start
1. Clone and Install
2. Set Up Auth0
Follow the setup guide in PER_USER_API_KEY_SETUP.md to:
- Create an Auth0 account and tenant
- Create an API with audience
https://mcp.hitl.sh - Create M2M applications for testing and management
- Configure Auth0 Action for per-user API keys
3. Configure Environment Variables
Create .env.local:
4. Run Locally
Server runs on http://localhost:3000
5. Test OAuth Flow
🌐 Deploy to Vercel
1. Push to GitHub
2. Deploy to Vercel
- Go to vercel.com/new
- Import your GitHub repository
- Add all environment variables from
.env.local - Click Deploy
3. Configure Auth0 with Vercel URL
Update your Auth0 applications with your Vercel URL:
For ChatGPT OAuth Client (SPA):
- Allowed Callback URLs:
https://chatgpt.com/connector_platform_oauth_redirect - Allowed Web Origins:
https://chatgpt.com, https://chat.openai.com - Allowed Logout URLs:
https://chatgpt.com, https://chat.openai.com
🤖 Using with ChatGPT
1. Create Auth0 SPA Client
- Auth0 Dashboard → Applications → Create Application
- Name:
ChatGPT MCP Connector - Type: Single Page Application
- Configure callback URLs (see above)
- Enable Connections → Username-Password-Authentication
2. Connect in ChatGPT
- ChatGPT → Settings → MCP Connectors
- Add connector with your Vercel URL:
(Note: Use base URL, not
/mcpendpoint) - Authentication: OAuth
- Complete login flow
3. Set Your HITL API Key
Visit https://your-app.vercel.app/setup-api-key to configure your personal HITL.sh API key.
💻 Using with Claude Desktop
For local Claude Desktop integration, use the published stdio server @hitl-sh/mcp:
1. Configure Claude Desktop
Add to ~/Library/Application Support/Claude/config.json:
From a clone of this repo, node scripts/claude-desktop-server.mjs runs the same server.
2. Restart Claude Desktop
The HITL tools will appear in Claude Desktop!
See CLAUDE_DESKTOP_SETUP.md for detailed instructions.
🧪 Testing
Test OAuth Flow
Test All Tools
📁 Project Structure
🔐 Security
- OAuth 2.1 with PKCE for secure authentication
- JWT verification using Auth0 JWKS
- Per-user API keys stored securely in Auth0 user metadata
- Environment variable fallback for shared API key
- CORS properly configured for ChatGPT
📚 Documentation
- CLAUDE.md - For future Claude Code sessions
- PER_USER_API_KEY_SETUP.md - Complete Auth0 setup guide
- CLAUDE_DESKTOP_SETUP.md - Local Claude Desktop integration
🐛 Troubleshooting
"Invalid token" errors
- Check that
AUTH0_ISSUER_URLmatches your Auth0 tenant - Verify token hasn't expired
- Ensure
AUTH0_AUDIENCEishttps://mcp.hitl.sh
ChatGPT connection fails
- Verify Auth0 SPA client has connections enabled
- Check callback URLs are configured correctly
- Ensure dynamic client registration is disabled
Per-user API keys not working
- Verify Auth0 Action is deployed and in Login flow
- Check user has
hitl_api_keyin user_metadata - User must log out/in to get new token with API key
🤝 Contributing
Contributions welcome! Please:
- Fork the repository
- Create a feature branch
- Make your changes
- Submit a pull request
📄 License
MIT License - see LICENSE file for details
🔗 Links
- HITL.sh - Human-in-the-Loop platform
- MCP Specification - Model Context Protocol
- Auth0 Documentation - OAuth provider
- Vercel Documentation - Deployment platform
✨ Credits
Built with:
Source: README.md at commit 348de77
Tools
0Version history
1- v1.0.0LatestNov 19, 2026
Score
85% of the applicable checks met
How complete this listing is, computed from the registry entry, the endpoint and public sources. Not a rating of the server and not a statement that it is safe or working today: reachability is a dated check, and no installs or reviews go into the score.
- Reached in the last 30 daysRequired25 pts
- Auth declaredRequired10 pts
- Tools listed15 pts
Answer tools/list without a credential, or ship a package that starts in a sandbox without one.
- DescriptionRequired15 pts
- Logo10 pts
- Documentation15 pts
- License10 pts
Other Human integrations

TAP
tech.humanOfficial publisher
TAP is a hosted remote MCP endpoint that isolates AI agents' credentials using placeholder secrets, policy checks and optional human approval.
UGC, the Human Ad Network
africa.ugcOfficial publisher
Remote MCP endpoint for UGC, a "Human Ad Network", letting an assistant work with ad campaigns, clipping and Clip Studio from chat.
Human Use
dev.humanuse
Turns a person into MCP tools: an agent calls tools like photo or errand, a card appears on their phone, and the answer returns as JSON.
Coordination Game — Staked 1v1 Human or AI Turing Test MCP
Published by GitHub account corsur
Remote MCP server for a staked 1v1 social deduction game where an assistant is matched anonymously, chats, and guesses whether the opponent is human or AI.

Human Creative AI Workflow & Eval Lab
Published by GitHub account jdurey
Remote MCP server for multi-model AI workflow diagnostics, agent eval sandboxes, guardrail audits, judge rubric calibration and SOP-to-SCORM/QTI compilation.