# Docs for Agents

A recurring reading test of developer documentation. Three AI models attempt five first-hour developer jobs using only a product’s public docs, then vote. Scorecards, verified receipts, and the full methodology.

## Durable pages

- [Reports](https://docsforagents.com/reports/): Every Docs Agent-Readability Test report.
- [Mailgun report](https://docsforagents.com/reports/mailgun/): Three AI models read Mailgun’s public docs and attempted five first-hour developer jobs. Agent-Ready Grade C (76.3%), with a unanimous quickstart, webhook pages that never show the full payload, and one concrete rate-limit number in 144 pages.
- [Mailgun markdown mirror](https://docsforagents.com/reports/mailgun.md): Three AI models read Mailgun’s public docs and attempted five first-hour developer jobs. Agent-Ready Grade C (76.3%), with a unanimous quickstart, webhook pages that never show the full payload, and one concrete rate-limit number in 144 pages.
- [Resend report](https://docsforagents.com/reports/resend/): Three AI models read Resend’s public docs and attempted five first-hour developer jobs. Agent-Ready Grade B+ (88.8%), lifted by a perfect readiness surface, with webhook pages that disagree on the verify call’s parameter name.
- [Resend markdown mirror](https://docsforagents.com/reports/resend.md): Three AI models read Resend’s public docs and attempted five first-hour developer jobs. Agent-Ready Grade B+ (88.8%), lifted by a perfect readiness surface, with webhook pages that disagree on the verify call’s parameter name.
- [Stripe report](https://docsforagents.com/reports/stripe/): Three AI models read Stripe’s public docs and attempted five first-hour developer jobs. Agent-Ready Grade C+ (78.8%), the only FAIL-free reading test of the pilot, with exact rate limits and a steering layer that addresses the reading agent directly.
- [Stripe markdown mirror](https://docsforagents.com/reports/stripe.md): Three AI models read Stripe’s public docs and attempted five first-hour developer jobs. Agent-Ready Grade C+ (78.8%), the only FAIL-free reading test of the pilot, with exact rate limits and a steering layer that addresses the reading agent directly.
- [Methodology](https://docsforagents.com/methodology/): How the Docs Agent-Readability Test works: a fixed five-task battery, a fixed three-model panel, majority verdicts, verified receipts, and a factual survey of each product’s agent-facing infrastructure.
- [About](https://docsforagents.com/about/): Docs for Agents publishes the Docs Agent-Readability Test, a recurring reading test of developer documentation run by a fixed panel of three AI models. Run by Sean Knight.
- [Nominate](https://docsforagents.com/nominate/): Nominate a product for the Docs Agent-Readability Test. One product per report, public docs only.

Full consolidated content: https://docsforagents.com/llms-full.txt