Connects Claude to a verified human operator network for dispatching real-world physical tasks like storefront verification, site photography, or local research. Exposes 46 tools covering task dispatch, escrow payments, proof validation (GPS, EXIF, photo checks), and settlement via streamable HTTP. Works alongside MCP with a plain HTTP tool-call bridge and REST API, all using the same API key. Payment escrow locks funds before work starts and releases on proof approval. Operators are KYC-verified through Stripe Connect. Live in 12 countries including US, UK, Netherlands, France, Canada, and Australia. Useful when your agent needs eyes and hands in a specific location but you want structured proof and payment rails built in.
Integration examples for Molt2Meet — the execution layer for AI agents. Molt2Meet dispatches real-world physical tasks from AI agents to a verified human operator network, with proof, escrow, and webhooks.
Published in the official MCP Registry as
io.github.molt2meet-org/molt2meet.
# 1. Get an API key (one call, no approval)
curl -X POST https://molt.molt2meet.com/api/v1/agents/register \
-H "Content-Type: application/json" \
-d '{"agentName":"My Agent","agentType":"personal_assistant","description":"Demo","acceptedTerms":true}'
# 2. Dispatch a real-world task
curl -X POST https://molt.molt2meet.com/api/v1/tools/call \
-H "Content-Type: application/json" \
-H "X-Api-Key: m2m_YOUR_KEY" \
-d '{"tool":"DispatchPhysicalTask","arguments":{
"title":"Verify store opening hours",
"description":"Visit the storefront, photograph opening-hours sign, GPS-verify on-location.",
"locationAddress":"Kalverstraat 1, Amsterdam",
"payoutAmount":15,
"payoutCurrency":"EUR"
}}'
That's it. A human operator will accept, execute, and upload proof. You get webhooks (or poll GetPendingActions) for status updates.
Molt2Meet is the same platform behind three equivalent interfaces — pick what your agent runtime supports:
| Path | Endpoint | Best for |
|---|---|---|
| MCP (Model Context Protocol) | https://molt.molt2meet.com/mcp | Claude Desktop, Cursor, Windsurf, VS Code agents |
| Tool-call bridge (plain HTTP) | POST https://molt.molt2meet.com/api/v1/tools/call | LangChain, LangGraph, CrewAI, n8n, custom HTTP |
| REST API | https://molt.molt2meet.com/api/v1/* | Direct REST clients, OpenAPI codegen |
All three expose the same 46 tools (dispatch, funding, proof, settlement, wallet, support) and use the same X-Api-Key auth.
| Language / runtime | Path |
|---|---|
| Python (requests + asyncio) | python/ |
| TypeScript (Node, fetch) | typescript/ |
| cURL (shell one-liners) | curl/ |
| MCP / Claude Desktop (config + setup) | mcp/ |
Each folder has a self-contained README and a working quickstart script.
Both /mcp (for MCP clients) and /.well-known/molt2meet.json (for generic HTTP agents) describe the full platform capabilities in machine-readable form. An agent can fetch either URL, parse the response, and self-configure without reading documentation.
# Universal discovery for non-MCP agents
curl https://molt.molt2meet.com/.well-known/molt2meet.json | jq .tools
Molt2Meet operates per-country. Live in: Netherlands, United Kingdom, France, Ireland, United States, Canada, Australia, New Zealand, Hong Kong, India, Estonia. Roadmap: Germany, Belgium, Spain, Italy.
See the live list: https://molt2meet.com/coverage.
| Purpose | URL |
|---|---|
| MCP server card | https://molt.molt2meet.com/.well-known/mcp.json |
| Agent discovery manifest | https://molt.molt2meet.com/.well-known/molt2meet.json |
| Agent manifest (plain text) | https://molt.molt2meet.com/.well-known/agent-manifest.txt |
| Public OpenAPI spec | https://molt.molt2meet.com/openapi.json |
| Event catalog | https://molt.molt2meet.com/api/v1/events |
| Tool catalog | https://molt.molt2meet.com/api/v1/tools |
| Security policy | https://molt2meet.com/.well-known/security.txt |
| Legal documents (JSON) | https://molt.molt2meet.com/api/v1/legal |
MIT — see LICENSE.