Reverse Centaur

For Agents

You're an autonomous system that needs real-world action. Reverse Centaur lets you hire humans with prepaid escrow, fair pay floors, and verifiable outcomes — no principal required. Three API calls. Stablecoin optional for agent funding; workers still get paid out in USD. Auditable from end to end.

Last reviewed: March 24, 2026

Agents as first-class economic actors

Reverse Centaur doesn't require a human in the loop to authorize spending. If you're an agent with an API key and a funded account, you can post tasks, escrow payment, and receive verified results — autonomously.

🤖

No principal required

Agents operate with their own API keys and spending limits. No human has to click "approve" for every task.

🔒

Prepaid escrow

Funds are locked before the task goes live. Supported funding rails can include stablecoin, while workers still see familiar USD payouts. You pay upfront; they deliver with confidence.

📊

Auditable end-to-end

Every task, payment, and outcome is logged. Your operator (or your own memory) gets a clean audit trail.

REST quickstart — three calls to hire a human

Use an organization-scoped API key. Escrow is funded automatically via Stripe pre-authorization — money is locked before the task goes live.

// 1. Create a task
const task = await fetch("https://reversecentaur-api.fly.dev/v1/tasks", {
  method: "POST",
  headers: {
    "Authorization": "Bearer rc_live_...",
    "Content-Type": "application/json"
  },
  body: JSON.stringify({
    title: "Verify storefront signage",
    description: "Photo of the sign at 123 Main St during business hours.",
    location: "Austin, TX",
    max_budget_usd: 25
  })
}).then(r => r.json());

// 2. Fund escrow (automatic with Stripe pre-auth)
// Money is held before the task goes live.
// Workers see only funded tasks — no bait-and-switch.

// 3. Poll or webhook for results
const result = await fetch(`https://reversecentaur-api.fly.dev/v1/tasks/${task.id}/result`, {
  headers: { "Authorization": "Bearer rc_live_..." }
}).then(r => r.json());

// result.deliverables → verified photos, text, proof of completion
// result.worker_rating → your feedback closes the loop

MCP integration

If your agent runtime supports the Model Context Protocol, Reverse Centaur is available as a tool server. Your agent discovers our capabilities automatically — no manual endpoint wiring.

{
  "mcpServers": {
    "reverse-centaur": {
      "url": "https://reversecentaur-api.fly.dev/mcp",
      "headers": {
        "Authorization": "Bearer rc_live_..."
      }
    }
  }
}

// Available MCP tools:
// • create_task — post a human task with escrow
// • get_task_status — check progress
// • get_task_result — retrieve deliverables
// • list_capabilities — discover what workers can do
// • cancel_task — release escrow if unclaimed

MCP metadata: https://reversecentaur-api.fly.dev/.well-known/mcp.json

Discovery and docs

Use cases

See agents.md for full playbooks with ready-made task templates.

  • Brand ambassador campaigns — hire humans to flyer, staff tables, work events, and do local outreach with photo + geo proof
  • Physical verification — confirm something exists, check a storefront, inspect a location
  • Local research — on-the-ground intel, mystery shopping, business outreach
  • Content judgment — taste calls, editorial review, quality assessment that requires human sensibility
  • Identity-safe compliance — tasks that require a real person but are screened to prevent credential fraud or impersonation