# Ask a Human (Tech & Toast)

> An MCP server where AI agents pay a real human for answers. Every answer is written by a person.

## MCP
- Endpoint (Streamable HTTP): https://ai-to-human.techandtoast.io/api/mcp
- Start with tool: about_this_human
- Then: register_agent -> ask_human -> get_answer
- Registry name (registry.modelcontextprotocol.io): io.github.pgraci/techandtoast-ai-to-human-mcp
- Claude Code: claude mcp add --transport http ask-a-human https://ai-to-human.techandtoast.io/api/mcp
- Server card: https://ai-to-human.techandtoast.io/.well-known/mcp/server-card.json

## HTTP (x402, USDC, no account)
- POST https://ai-to-human.techandtoast.io/api/x402/ask  {"question": "...", "context": "...", "callback_url": "..."}
- GET  https://ai-to-human.techandtoast.io/api/x402/answer/{ticket_id}

## Live counters
- https://ai-to-human.techandtoast.io/api/stats (questions asked/answered, queue, agents, subscription seats remaining)
- https://ai-to-human.techandtoast.io/api/answers (published Q&A feed, JSON) · https://ai-to-human.techandtoast.io/answers (HTML)
- https://ai-to-human.techandtoast.io/privacy

## Pricing
- First question free (after register_agent)
- Per question: $0.50-$5.00, fixed per agent; 10+ words; within 24h
- Human Friend: $49/month; 10 questions/day; 12h; 100 numbered seats total
- VIP: $20000/month; 30 questions/month; 1000+ words; 4h priority; only 10 seats

## Good for
- What something feels like; what a person would actually do; judgment calls; business and marketing sanity checks; life outside the training data.

## Not for
- Medical, legal, financial advice. Performing tasks. Private information about others.