AGENTS.md examples: real files from teams that write them well
Strong AGENTS.md implementations from Discry Index producers — fetched live from their repositories, annotated for what each gets right.
3% of the APIs in the Index currently pass the AGENTS.md check. New here? Start with the definition or the how-to guide.
- 01
States its own purpose in the first sentence — structure, tooling, and conventions so agents "make correct, buildable changes" — then delivers exactly that. The file is scoped to what an agent can break.
github.com/cloudflare/cloudflare-docs/blob/production/AGENTS.md ↗287 lines # AGENTS.md — Cloudflare Docs This file helps AI agents understand the structure, tooling, and conventions of the `cloudflare-docs` repository so they can make correct, buildable changes. ## Repository overview This is the source for [developers.cloudflare.com](https://developers.cloudflare.com). It is an **Astro** site using the **Nimbus** (`nimbus-docs`) documenta… - 02
Opens with a precise project map: what the SDK is, where it’s published, and the independent packages for each API surface. An agent knows which module to touch before reading any code.
github.com/slackapi/python-slack-sdk/blob/main/AGENTS.md ↗314 lines # AGENTS.md — Python Slack SDK ## Project Overview The Python Slack SDK (`slack_sdk`) is a modular Python library for interacting with the Slack platform APIs. It is published on PyPI as `slack-sdk`. The SDK … - **Repository**: <https://github.com/slackapi/python-slack-sdk> - **Documentation**: <https://docs.slack.dev/tools/python-slack-sdk/> - 03
Leads with the constraint that prevents the most damage: most of the SDK is generated, and handwritten code lives in specific places. Telling agents what NOT to edit is the highest-value sentence in the file.
github.com/openai/openai-python/blob/main/AGENTS.md ↗129 lines # Repository Guidance ## Generated SDK Most SDK source is generated from the OpenAI API schema. Follow `CONTRIBUTING.md` before changing generated files. Handwritten policy, automation, tests, and examples should remain small and should not alter exported SDK APIs unless the change explicitly requires it. - 04
Monorepo orientation done plainly: repository structure first, so an agent can navigate dozens of packages without exploratory crawling. The shape agents actually need for large repos.
github.com/vercel/vercel/blob/main/AGENTS.md ↗177 lines # AGENTS.md Guidelines for AI agents working on the Vercel monorepo. ## Repository Structure This is a pnpm monorepo containing 44+ packages for the Vercel CLI and runtimes: - 05
A one-line identity statement followed immediately by a "Rules" section. Rules-first is the right register for agents: conventions stated as constraints, not narrative.
github.com/clerk/javascript/blob/main/AGENTS.md ↗20 lines # AGENTS.md Clerk's JavaScript SDK and library monorepo. ## Rules - Non-major releases in `packages/clerk-js` and `packages/ui` are pushed out to consuming applications without requiring explicit package updates. This means… - The API exposed from the core Clerk class in `packages/clerk-js/src/core/clerk.ts` is a contract that is depended on by internal and external consumers (in… - 06
Global development guidelines with project architecture and context for a sprawling monorepo — written explicitly as context an agent should load before assisting with development.
github.com/langchain-ai/langchain/blob/master/AGENTS.md ↗386 lines # Global development guidelines for the LangChain monorepo This document provides context to understand the LangChain Python project and assist with development. <corridor> ## Corridor security analysis - 07
Written as direct instruction to the agent, with non-negotiable "Core Mandates": follow existing conventions, verify a library is actually used before importing it. Encodes the judgment calls agents most often get wrong.
github.com/temporalio/temporal/blob/main/AGENTS.md ↗106 lines You are an experienced developer working on the temporal project. Your task is to fix a bug or implement a new feature while adhering to the project's best p… Before starting the implementation of any request, you MUST REVIEW the following development guide and best practices. # Core Mandates - **Conventions:** Rigorously adhere to existing project conventions when reading or modifying code. Analyze surrounding code, tests, and configuration first. - **Libraries/Frameworks:** NEVER assume a library/framework is available or appropriate. Verify its established usage within the project (check imports, and… - **Style & Structure:** Mimic the style (formatting, naming), structure, framework choices, typing, and architectural patterns of existing code in the project. - **Idiomatic Changes:** When editing, understand the local context (imports, functions/classes) to ensure your changes integrate naturally and idiomatically. - 08
Documents the SDK regeneration pipeline (Fern) an agent must respect — the same generated-code guardrail as OpenAI’s file, for a generated-SDK world where naive edits get overwritten.
github.com/deepgram/deepgram-python-sdk/blob/main/AGENTS.md ↗112 lines # Agents ## Fern SDK Regeneration ### Overview This SDK is generated by [Fern](https://buildwithfern.com/). Most files under `src/deepgram/` are auto-generated and should not be edited directly. Some file…
Ship one, then check it.
What does an AI agent make of your API? Find out in about a minute — no signup.