Free browser utility

AGENTS.md generator

Write repository instructions from the facts you know. Draft a concise AGENTS.md with your setup and verification commands, edit boundaries, conventions and handoff expectations.

This utility generates in your browser without uploading input. No signup or AI provider is needed. Use non-secret project context.

Project purpose (required)

Explain what the project does and who it serves. This is the only required field. Limit: 2,000 characters.

Setup commands

Enter the exact commands and working directory needed to prepare this project. Nothing will be run or inferred. Limit: 3,000 characters.

Verification commands

Enter checks an agent should run and their working directory. These are instructions, not evidence of a completed run. Limit: 3,000 characters.

Edit boundaries

Name allowed paths and protected or generated files. An empty field does not grant permission to edit everything. Limit: 3,000 characters.

Conventions

Add project-specific naming, architecture or style rules that an agent needs. Prefer short, concrete instructions. Limit: 3,000 characters.

Handoff evidence

Specify what the agent should report when finished, such as changed files, commands run, observed results and untested gaps. Limit: 3,000 characters.

Share warning: your inputs appear in the link, readable by anyone who receives it. Do not include secrets or confidential information. Copying does not send the link or change this page address.

Generated AGENTS.md

Start with the example or enter your own facts. Use Generate AGENTS.md when you are ready, then review the result.

Methodology and format

Enter the project purpose, then supply only the commands and instructions you have checked. Generate AGENTS.md formats those inputs in a fixed order. It does not inspect a repository, execute commands or call an AI provider.

AGENTS.md is a Markdown file for repository guidance. This generator uses six practical sections rather than a required schema. Blank optional sections are omitted; no setup command, test command, edit permission or convention is inferred. The purpose requirement belongs to this form, not the AGENTS.md format.

The export uses fixed headings and fenced text blocks for every supplied instruction and command. Input text is preserved literally inside a fence longer than any supplied backtick run, so its Markdown punctuation cannot create new document structure. Review the plain-text file before adding it to your repository.

Each input is bounded to the displayed limit: 2,000 characters for purpose and 3,000 for each optional field, counted as UTF-16 code units. Surrounding whitespace is trimmed. Imported drafts accept only known fields and reject malformed values. Share URLs are limited to 8,000 characters; use a file export for larger drafts.

This utility generates the file in your browser without uploading input or automatically saving drafts. Copy, download and sharing happen on request. Copy share link includes your inputs in the URL fragment, readable by anyone with the link. Use non-secret drafting context. Other website features have their own data handling described in the privacy policy.

AGENTS.md format guide — Describes repository guidance in standard Markdown, optional headings and repository-root or nested instruction files. Agent-specific behavior still needs checking.

CommonMark: fenced code blocks — Documents the literal fenced blocks used for supplied commands in the Markdown export.

Limitations

The output is a repository instruction draft, not a universal safety policy, permission system or guarantee that any agent follows it. Check how your chosen coding agent loads repository instructions and resolves conflicts.

Listed verification commands are requested checks, not proof that they ran or passed. The generator cannot validate commands, detect conflicting rules, inspect protected paths or determine whether your instructions are current.

The trail catalog example is fictional. Adapt its commands and paths to your own repository; they have not been run by this utility. Review existing instructions before replacing a file, and keep credentials and confidential details out of shared drafts.

Frequently asked questions

What is an AGENTS.md file?

AGENTS.md is a Markdown file that gives coding agents repository context and instructions, such as project purpose, setup steps, verification commands and conventions. It has no required set of headings; this generator offers a concise starting structure.

Does the generator discover my project's commands?

No. It formats only the text you supply and omits blank optional sections. It does not inspect files, infer package-manager commands or run checks. A verification command in the output is an instruction, not a successful test result.

Where should I put the generated AGENTS.md?

Download the file, review it against existing repository guidance, and add it where your coding agent expects project instructions. The AGENTS.md guide describes a repository-root file and nested files for subprojects. Check your agent's loading and precedence rules before adopting it.

Is my project text uploaded by this utility?

Generation runs in your browser without uploading input or automatically saving drafts. Copy and download are explicit actions. A share link includes your inputs in its URL fragment and is readable by anyone with the link. Use non-secret context and see the privacy policy for other website features.

Related guides and tools

Build task-specific context for coding handoffs

Record bounded handoff events with Claude Code hooks

Draft a one-session agent handoff

Coordinate work across coding agents

Multi-agent orchestration with Claude Code, Cursor and Codex

Website privacy policy