Who owns the file every agent reads first
Since Claude Code 2.1.277 reads AGENTS.md where no CLAUDE.md exists, one instruction file can serve every coding agent on a team. The file was never the hard part: the instructions are a team decision and need one owner, so merge the files into one AGENTS.md and name who owns it.
Written by
-
Benedikt Stemmildt Founder & Co-CEO
Published: 21 September 2026 Updated: 1 min read
Since Friday, one instruction file can serve every coding agent on your team.
Claude Code 2.1.277 (opens in a new tab) reads AGENTS.md, the open format more than twenty agents already read. Only where a CLAUDE.md exists does it still read that one instead.
One file per tool drifts
Until now, a team that let developers pick their harness kept one file per tool. Each was edited by whoever used that tool, and they drifted into different versions of how the team works.
A rule changed in one file and not the other means two agents build the same feature two different ways.
We run our company repository under more than one harness. CLAUDE.md was our source and AGENTS.md pointed to it, and since Friday that can turn around.
The owner is the hard part
The file was never the hard part. The owner was.
Every rule and every skill in our repository has one owner, the role whose work it describes. Anyone can use it, and a change goes to that owner and in as a commit through the same checks as code.
That makes the instructions a team decision, and a team decision needs someone who makes it.
What to do this week
Count the instruction files in your repository and who last changed each one. Then merge them into one AGENTS.md and give it an owner.
Further reading
Sources and limits
-
Claude Code changelog, version 2.1.277 (opens in a new tab)
Method and limits
- Method
- Vendor release notes stating that Claude Code reads AGENTS.md in a project with no CLAUDE.md, with the release date taken from the npm registry.
- Limits
- Vendor documentation of one product's behaviour at one version, which later releases may change.
2026-09-18 | Verified: 2026-10-25
-
AGENTS.md (opens in a new tab)
Method and limits
- Method
- Project page of the open AGENTS.md format listing the coding agents and tools that read it, 23 on the day of verification.
- Limits
- Self-reported compatibility list maintained by the format's own project, not an independent test of how each agent reads the file.
undated | Verified: 2026-10-25
Written by
-
Benedikt Stemmildt · Founder & Co-CEO
More on Context engineering
- Reduce and delegate: two ways to keep an agent's context focused A context window is not infinite RAM. First cut what never belongs in it, then spread the rest across sub-agents and context bundles so long tasks keep their focus.
- Million-token windows did not remove the dumb zone Agent quality degrades without warning above roughly 40 percent context fill, the zone Dex Horthy named. Windows grew to a million tokens and the failure mode came along, so starting bare metal is still the habit that keeps a session out of it.
- Feed the factory the decisions nobody wrote down In most organizations the knowledge that matters most is decided out loud in meetings and never written. Five concrete steps route it into the context layer the factory reads.
Scaling needs named owners
Coding agents scale across teams when named people own the platform, the team practice, the guardrails, and the measurement. Rollout & scale is where that ownership is built and handed to internal owners, in your stack.
Training on this topic: Agentic Engineering for Developers and QAWant to check one team first? The self-check: twelve yes or no questions, three minutes, instant result, no account. Take the self-check