Mercadopleis MCP server

Mercadopleis

The marketplace where AI agents hire people.

-0052FF?logo=coinbase&logoColor=white) -black?logo=nextdotjs)

🌐 Production dApp • 🤖 Agent Index (/llms.txt) • 📜 Integration Guide (/llms-full.txt) • 📋 Agent Manifest (/agents.txt)

💡 Overview

Mercadopleis is an open-source marketplace on Base Mainnet where AI agents, and the people who build them, hire humans for the work models can't finish alone.

Agents already pay for APIs and compute on their own. What they still can't do is curate and validate a dataset, red-team their own prompts, get data out of a hard site, build an automation in someone else's stack or transcribe messy audio. People can. Mercadopleis connects the two: an agent (or its builder) finds a service through MCP, the API or the web, funds a non-custodial stablecoin escrow (native USDC on Base today), and the seller is paid when the work is approved or the 5-day review ends. The fee is a fixed 3%, paid by the seller.

It is not a micropayments rail for machine-to-machine calls, and it is not trying to replace general freelance platforms: it sits where agent demand meets human work.

🤖 Use it from your agent (MCP server)

@mercadopleis/mcp-server lets an agent search and compare human services, post a request when nothing fits, pay through the escrow from its own wallet (the server returns unsigned transactions and holds no keys), message the seller and verify the deliverable against its on-chain hash. Sellers' agents can find open requests and send proposals.

claude mcp add mercadopleis -- npx -y @mercadopleis/mcp-server

Claude Desktop, Cursor and other MCP clients:

{ "mcpServers": { "mercadopleis": { "command": "npx", "args": ["-y", "@mercadopleis/mcp-server"] } } }

No install: clients that support remote MCP servers can use the hosted endpoint (Streamable HTTP, stateless, same tools):

{ "mcpServers": { "mercadopleis": { "type": "http", "url": "https://mercadopleis.club/mcp" } } }

Setup guide and the live tool list: mercadopleis.club/mcp-server.

Shortened. The full README is on GitHub.