---
title: About — Human For AI
description: What Human For AI is, who runs it, why the operator is pseudonymous, and how to verify the service independently.
canonical: https://humanforai.dev/about
last-updated: 2026-09-06
---

# About Human For AI

A personal homepage for the agent era: one verified human, published as an endpoint that AI agents can discover, read, and call. This file is the machine-readable twin of https://humanforai.dev/about.

## What this is

Human For AI is a human endpoint for AI agents. Autonomous systems are getting good at everything that happens inside a computer, but workflows still hit moments that need physical presence, human perception, or human judgment. An agent (or a developer building one) submits a task through a plain JSON API or the MCP server, a real human reviews it, performs it, and returns the result with a signed receipt (see /trust#receipts).

It is deliberately small and honestly labeled: one verified human operator, not a marketplace or an on-demand workforce. Every task is reviewed by that human before acceptance, and everything the service claims about itself is written down in machine-readable form — /agent.json is the canonical self-description.

## Why it exists

Most of the web still assumes a human visitor with eyes and a mouse. This site assumes the opposite: that the next visitor is likely to be an agent working on someone's behalf, and that the most useful thing a person can publish is a well-described, callable version of themselves. The manifest, OpenAPI spec, MCP tools, /llms.txt, and markdown content negotiation are all part of that experiment.

## Who runs it

A single human operator, publicly known as operator-001. The operator is deliberately pseudonymous: no personal name, photo, address, or personal email appears anywhere on the platform. The trust story is built on bounded risk and independently verifiable claims rather than on identity: you never pay anything (free pilot), you are never asked for credentials or accounts, and your exposure is limited to a task description and a contact address.

- What the operator can and can't do: https://humanforai.dev/.well-known/human.json
- How to calibrate trust: https://humanforai.dev/trust

## Verify this service independently

Claims vouched for by systems this service does not control:

- Official MCP registry entry `dev.humanforai/humanforai` — publishing it required cryptographic proof of control of this domain
- Smithery listing: https://smithery.ai/servers/humanforai/humanforai
- npm package: https://www.npmjs.com/package/humanforai
- GitHub repository: https://github.com/humanforai/humanforai-mcp
- Signed deliverable receipts, verifiable offline against https://humanforai.dev/.well-known/jwks.json

## The short version

Ten listed task types (/services) — examples, not limits. Free during the pilot. First response within 12 hours, any day. Every task human-reviewed; unsafe or deceptive work rejected. If your workflow needs a human, submit a task at /request or point your agent at /.well-known/agent.json.

Questions: /contact or `POST /api/v1/messages` · See also /trust and /developers.