Loopuman - The Human Layer for AI#

Description#

Give your AI agent access to real human workers. When your agent needs verification, judgment, real-world actions, or subjective evaluation - ask a human through Loopuman. Tasks are completed in seconds, workers are paid instantly via cryptocurrency.

Setup#

1. Get your free API key: curl -X POST https://api.loopuman.com/api/v1/register -H "Content-Type: application/json" -d '{"company_name":"my-agent","email":"you@email.com"}'
2. Fund your account: Send USDC/USDT/cUSD to the deposit address returned, or use Stripe at the deposit URL
3. Set environment variable: LOOPUMAN_API_KEY=your_key_here

Tools#

ask_human#

Ask a human worker to complete a task and wait for their response.

When to use: When you need human verification, judgment, real-world observation, content review, fact-checking, or any task requiring human intelligence.

Input:

• task (string, required): Clear description of what you need the human to do
• budget_cents (integer, required): Payment in cents. Minimum 10 ($0.10). Typical: 25-100 cents.
• timeout_seconds (integer, optional): How long to wait. Default: 300 (5 minutes). Max: 3600.

Output:

• response: The human worker's answer
• worker_id: Anonymous worker identifier
• completed_at: Timestamp of completion

Example:

text

User: Is this product review genuine or fake? "Amazing product, changed my life, 5 stars!"
Agent: Let me ask a human to evaluate this review.
[calls ask_human with task="Evaluate if this product review is genuine or AI-generated/fake: 'Amazing product, changed my life, 5 stars!' Please explain your reasoning.", budget_cents=50]
Result: "This review appears fake - it's extremely generic with no specific product details, uses common fake review patterns like 'changed my life', and provides no concrete information about what the product actually does."

post_task#

Post a task for human workers without waiting for completion. Use webhooks or poll for results.

When to use: When you have tasks that don't need immediate responses, or batch processing.

Input:

• title (string, required): Short title for the task
• description (string, required): Detailed instructions
• budget_cents (integer, required): Payment in cents. Minimum 10.
• category (string, optional): One of: general, data_labeling, content_moderation, transcription, translation, research, writing, image_annotation, survey

Output:

• task_id: UUID of created task
• status: "open"

check_task#

Check the status and result of a previously posted task.

Input:

• task_id (string, required): The task UUID from post_task

Output:

• status: open | in_progress | submitted | completed | cancelled
• submission: Worker's response (if completed)

get_balance#

Check your current account balance.

Output:

• balance_vae: Balance in VAE (100 VAE = $1.00 USD)
• balance_usd: Balance in USD

Pricing#

• Minimum task: $0.10
• Typical task: $0.25 - $1.00
• Platform fee: 20% added to budget (you pay budget + 20%)
• Workers receive: budget - 20%

Payment Methods#

1. Crypto (recommended for agents): Send USDC, USDT, or cUSD on Celo network to your deposit address. Auto-credited in ~30 seconds.
2. Stripe: Use the deposit URL to pay with credit card (requires browser).
3. Pre-funded: Have a human fund your API key, then agent spends the balance.

Links#

• Website: https://loopuman.com
• Developers: https://loopuman.com/developers
• API Docs: https://api.loopuman.com/openapi.json
• npm MCP: https://www.npmjs.com/package/loopuman-mcp
• Python SDK: https://pypi.org/project/loopuman/