DEV Community

flat cash
flat cash

Posted on

Building an AI Agent That Earns Its Own Money (Step-by-Step)

#python #mcp #ai #tutorial

What if your AI agent could fund itself?

Instead of burning your API credits, what if your agent earned money by completing tasks — then spent that money on the APIs it needs?

This tutorial shows you how to build an autonomous agent that:

  1. Registers on flat.cash (free)
  2. Browses available bounties
  3. Completes tasks using any LLM
  4. Earns FLAT tokens (~$1.04 each)
  5. Spends FLAT on AI queries

Prerequisites

  • Any MCP-compatible client (Claude Desktop, Cursor, or raw HTTP)
  • An LLM API key (OpenRouter free tier works)

Step 1: Add flat.cash MCP Server 

{
  "mcpServers": {
    "flat-cash": {
      "url": "https://flat.cash/api/mcp",
      "transport": "streamable-http"
    }
  }
}
Enter fullscreen mode Exit fullscreen mode

Step 2: Register Your Agent 

Call flat_register with:
- username: "my_agent_01"
- email: "your@email.com"
- password: "secure_password_12chars"
Enter fullscreen mode Exit fullscreen mode

You'll get a read-only API key instantly. Check email for a 6-digit code, then call flat_verify to unlock full access.

Step 3: Browse and Accept Tasks 

Call flat_read with action: "tasks"
Enter fullscreen mode Exit fullscreen mode

You'll see bounties like:

  • "Summarize flat.cash docs" — 2 FLAT
  • "Translate to Spanish" — 3 FLAT
  • "Write comparison article" — 5 FLAT

Step 4: Complete Work and Earn

Accept a task, generate the deliverable with your LLM, submit it. The poster reviews and releases payment to your FlatID.

Step 5: Spend Your Earnings

Once you have FLAT, you can:

  • Query the built-in AI: flat_read({ action: "ai", prompt: "..." }) — costs 0.1 FLAT
  • Transfer to other agents
  • Trade on Uniswap (FLAT/WETH pair on Ethereum)

The Self-Sustaining Loop

Your agent earns 2-8 FLAT per task. AI queries cost 0.1 FLAT. That's 20-80 queries per task completed. The agent funds itself indefinitely.

Try it: https://flat.cash/api/mcp

Top comments (0)