How we source and rank.

Methodology

AgentBuilt documents real microbusinesses that AI agents built or run. Every case follows the same 12-part template, so you can compare them side by side.

What gets included?

Businesses or experiments where an AI agent (Grok Bot, OpenClaw, Claude, Claude Code, Devin, GPT and others) did meaningful build or operating work, and where there's a public source we can link to. Experiments that failed are included on purpose: $0 results teach as much as wins.

How are numbers sourced?

• Every figure comes from a public post, interview, dashboard or article, and is cited inline as [n].
• We never estimate or invent revenue. Anything missing is marked "not disclosed".
• Self-reported figures aren't audited. We separate investment or funding from revenue.

The case template

1. TL;DR and key numbers
2. What it sells, and to whom
3. How it was built
4. Business model and pricing
5. Distribution and traffic
6. Revenue receipts and timeline
7. Agent vs. human split
8. Costs and unit economics
9. What worked and what didn't
10. Replication playbook
11. Risks and moat
12. FAQ and sources

Corrections

Spot an error? Submit a correction or DM @stauch1999.