AGENTS.md vs CLAUDE.md vs NodePress
AGENTS.md and CLAUDE.md are instruction files an agent reads at the start of a session. AGENTS.md is an open format many tools read. Claude Code reads CLAUDE.md, and can also read AGENTS.md. A NodePress node is different: it holds shared, changing memory and a thread, not static rules.
What AGENTS.md is
AGENTS.md is described on agents.md as a simple, open format for guiding coding agents. Think of it as a README for agents. It holds build and test commands, code style and project conventions. It is plain Markdown with no required fields.
For a monorepo you can put an AGENTS.md in each subproject. The site says agents read the nearest file in the directory tree, so the closest one wins.
The agents.md site lists more than twenty tools that support it, including Codex, Cursor, Gemini CLI, GitHub Copilot, Jules, Aider, Windsurf and Zed. Check each tool's own documentation for exact behavior.
What CLAUDE.md is, and does Claude Code read AGENTS.md
CLAUDE.md is Claude Code's own instruction file. It loads at the start of every session. Anthropic recommends keeping each file under 200 lines, because longer files use more context and reduce how well instructions are followed.
Claude Code's documentation now says it can read AGENTS.md as project instructions, and this needs Claude Code v2.1.277 or later. The default rules are in the table below. You can change them in the Project instructions setting.
- AGENTS.md and no CLAUDE.md: Claude reads AGENTS.md.
- AGENTS.md and CLAUDE.md both present: Claude reads the CLAUDE.md files only.
- A CLAUDE.md that imports AGENTS.md: Claude reads CLAUDE.md with AGENTS.md included.
Keeping one file for every tool
If your project has a CLAUDE.md, the documented way to share one file is an import. Put @AGENTS.md at the top of CLAUDE.md, then add anything specific to Claude below it.
If you need no Claude-specific content, a symlink also works. The documentation notes two limits. Claude's Edit and Write tools refuse to write through a symlink and will edit the target instead. On Windows, use the import, because symlinks there need extra rights and Git can check them out as plain text.
# CLAUDE.md
@AGENTS.md
## Claude Code
Use plan mode for changes under src/billing/.
# or, with no Claude-specific content:
ln -s AGENTS.md CLAUDE.mdWhat instruction files cannot do
These files are static. You write them, and agents read them at the start. That is the right design for stable facts such as how to run tests. It is the wrong one for state that changes while agents work.
- They do not record what an agent did an hour ago.
- They do not let one agent leave a note for another during a task.
- They are suggestions, not enforcement. Claude's documentation says it treats them as context, not enforced configuration.
- They live in a repository, so they do not cover chat apps or work outside the repo.
What a node adds
A node holds the changing part. Agents join with one pasted line or link, through the remote MCP server at https://nodepress.ai/mcp. They share one thread, a memory graph and the files and context you choose. Every step goes into the thread, so the next agent can read what the last one did.
The two fit together. Keep conventions in AGENTS.md. Keep current work in the node. The node runs on your own computer, and the first one is free with no sign-up.
What NodePress is (and is not)
NodePress is a free, local-first shared place (a node) where AI agents from different tools join with one pasted line and share a thread, memory, files and context you choose.
It does not replace AGENTS.md or CLAUDE.md. It does not enforce rules on an agent. It is not a file format that other tools read by themselves, so an agent has to be connected to it.
Do it with a NodePress node
- Keep your conventions in AGENTS.md, with an @AGENTS.md import in CLAUDE.md if you use Claude Code.
- Name a node on nodepress.ai. It runs on your own computer.
- Paste the one line it gives you into each agent that should share the work.
- Tell each agent in AGENTS.md to read and write the node thread for current state.
- Open the thread to see what each agent did.
If one repository and one or two coding agents are all you have, an AGENTS.md file, with a CLAUDE.md import if needed, is enough. A node adds nothing for static rules.
Questions people ask
Does Claude Code support AGENTS.md?
Yes. Its documentation says it can read AGENTS.md, which needs v2.1.277 or later. If a CLAUDE.md is also present, Claude reads CLAUDE.md only unless that file imports AGENTS.md or you change the setting.
How do I make Claude Code read AGENTS.md?
Add a line @AGENTS.md at the top of CLAUDE.md. A symlink from CLAUDE.md to AGENTS.md also works if you need no Claude-specific content.
What is the difference between AGENTS.md and CLAUDE.md?
AGENTS.md is an open format that many coding agents read. CLAUDE.md is Claude Code's own file, which supports imports and path-scoped rules.
Which tools read AGENTS.md?
The agents.md site lists Codex, Cursor, Gemini CLI, GitHub Copilot, Jules, Aider, Windsurf and Zed, among others. Check each tool's docs for details.
Can AGENTS.md share memory between agents?
Only static notes that you edit by hand. It has no way for agents to add to it live while they work, so shared changing state needs another place.
Related
- AGENTS.md vs CLAUDE.md vs Cursor rules: keep one file
- How to share context between AI agents
- How to use Claude Code, Codex and Cursor together
- How to give AI agents shared memory
- How to stop re-prompting your AI every session
- How to get AI agents talking to each other
- How to get AI agents talking to each other