Skip to content
How it learnsFrom coding sessions, reviews, Slack and docs.Only what fitsEach session gets the few rules its task needs.Pull request reviewEvery PR checked against your team’s rules.Keep it trueA gate, conflicts, disputes and stale rules, handled.HandoffsCarry the work to another session or a teammate.SecurityWhat it sends, what it changes and how to leave.
Engineering teamsOne agent or several: what one person learns, every session uses.PR review onlyInstall the GitHub App. Nothing on anyone’s machine.Teams that mix agentsClaude Code, Cursor, Codex and Copilot, the same rules.On your ownYour corrections follow you across repos, machines and agents.Book a call30 minutes with Roei, if you want a hand. Never required.
DocsPricingSign in
Connect your agents →
For teamsOn your ownPR review onlyHow it learnsOnly what fitsPull request reviewKeep it trueSecurityDocsPricingSign inConnect your agents →
Home/Blog

Blog

AGENTS.md, Cursor rules and Copilot instructions

  • AGENTS.md vs CLAUDE.md
  • Sync rules across repos
  • Why Codex ignores it
  • Keep rule files in sync
  • 3,905 files checked
  • Why Copilot ignores it
  • Cursor Team Rules
  • Cursor rules
  • copilot-instructions.md
  • Claude Code and AGENTS.md
  • AGENTS.md research

CLAUDE.md and context engineering

  • Why Claude ignores it
  • Test a CLAUDE.md rule
  • One CLAUDE.md for all projects
  • Context engineering, Claude 5
  • CLAUDE.md length
  • Which rules got used
  • Context rot, stale rules
  • Growing CLAUDE.md
  • Claude Code best practices
  • Karpathy CLAUDE.md
  • Committing CLAUDE.md

Built-in agent memory

  • Agent memory compared
  • Claude Code auto memory
  • GitHub Copilot Memory
  • Make Claude Code remember
  • Claude Tag vs Claude Code
  • Codex memories
  • Cursor memories
  • Memory or documentation?

Team knowledge and shared conventions

  • Claude Code team memory
  • Who owns CLAUDE.md
  • Claude Code second brain
  • Shared memory across agents
  • Stop repeated mistakes
  • Hand off between agents
  • Company brain
  • Slack decisions
  • Memory poisoning
  • Karpathy's LLM wiki
  • ADRs for agents

Skills, hooks and permissions

  • Rules vs skills
  • Claude Code hooks
  • Codex skills
  • Auto mode
  • CLAUDE.md vs skills vs hooks
  • Claude Code skills

AI code review and PR guardrails

  • PR review against AGENTS.md
  • Review comments to rules

Memory tools and alternatives

  • Best company brain tools
  • Unblocked alternatives
  • claude-mem alternatives
  • Supermemory alternatives
  • Mem0 alternatives
  • claude-mem
  • Best MCP memory servers
  • Supermemory Company Brain alternatives

Blog

Who should own CLAUDE.md and AGENTS.md on a team?

Published November 18, 2026

Give the rule files code owners. A CODEOWNERS entry for CLAUDE.md, AGENTS.md, .claude/, .cursor/rules/ and Copilot's instructions, plus GitHub's “Require review from Code Owners”, puts a named reviewer on every change to what the agents read. Then give that owner a regular prune, because nothing in these files expires.

That is the short answer. The rest of this note gives the file, what the owner should ask of each change, and what review alone does not catch. GitHub and Anthropic facts are from their docs, read on 18 November 2026.

Why a rules file needs an owner

A line in CLAUDE.md changes what every teammate's agent does on every task. It is often added inside a feature pull request, where the reviewer is reading the feature. Anthropic already asks for upkeep: the memory docs say to review CLAUDE.md files, nested ones and .claude/rules/ “periodically to remove outdated or conflicting instructions.” Without a name next to the file, periodically means never.

The repos that keep two rule files show what an unowned pair looks like. In our scan of public GitHub repos where a root CLAUDE.md sits beside a root AGENTS.md (the study, data):

What CLAUDE.md does with the AGENTS.md beside itShare of pairs
Never mentions it, so by default Claude Code never reads it16.4%
Asks Claude in a sentence to read it, so only if Claude opens it14.9%
Was copied from it once, then drifted7.0%
Copies its lines, a second copy to keep current6.6%

Each row is a fix a reviewer who owns both files would ask for in one comment: an @AGENTS.md import in place of the sentence or the copy.

The CODEOWNERS file

GitHub reads CODEOWNERS from .github/, the repo root or docs/, and “code owners are automatically requested for review when someone opens a pull request that modifies code that they own” (GitHub docs):

# .github/CODEOWNERS
# The last matching pattern wins, so keep these lines at the end.
CLAUDE.md                          @acme/agent-rules
AGENTS.md                          @acme/agent-rules
/.claude/                          @acme/agent-rules
/.cursor/rules/                    @acme/agent-rules
/.github/copilot-instructions.md   @acme/agent-rules
/.github/instructions/             @acme/agent-rules

A pattern with no slash, like CLAUDE.md, matches the file in any folder, so nested files are covered too. Order matters: “the last matching pattern takes the most precedence”, so a broad line such as * @acme/everyone placed after these would take them back. A request is not a requirement, though. To block a merge until the owner approves, turn on “Require review from Code Owners” in the branch protection rule or ruleset for your default branch.

Pick the owner by who answers for agent behaviour, not by who wrote the first file. On a small team that is one senior engineer; on a larger one, a group of two or three, so reviews do not wait on one person's holiday.

What the owner asks of each change

The changeThe question
A new ruleCould the agent learn this from the code? If yes, it costs context and adds nothing
A new ruleCan someone check it was followed? "Run pnpm test before committing", not "test your changes"
A new ruleMust it hold every time? Then it belongs in a hook or a deny rule, not prose
A rule that names a path, flag or toolDoes that thing still exist on the default branch?
Text pasted into a second fileCan it be an @import instead of a copy?
AnythingDoes it contradict a line in another rules file? Claude may pick either one
AnythingIs the file still under about 200 lines, Anthropic’s target?

The fuller version of the size and hook questions is in how long should CLAUDE.md be and CLAUDE.md vs skills vs hooks. To test whether a proposed rule changes anything before you approve it, see how to test a CLAUDE.md rule.

The prune: a recurring job, not a review

Review catches what changes. It does not catch the rule that went stale because the code moved under it, since nobody opened a pull request on the rules file. That needs a scheduled pass by the owner. Claude Code ships one: /doctor prompt-audit looks for instructions written for older models, references to files or commands that no longer exist, and files that contradict each other, and changes nothing until you ask it to apply the edits. Our free rules file checker says which agents load a file from where it sits. What goes stale, and why, is in context rot and stale rules.

Who owns the files outside the repo

  • A managed CLAUDE.md is owned by whoever ships it to the machines, usually IT through MDM. Developers cannot exclude it with claudeMdExcludes, so changes to it deserve the same review as the repo file.
  • Cursor Team Rules live in the Cursor dashboard on Team and Enterprise plans. An enforced rule “is required for all team members and cannot be disabled” (Cursor docs), and it has no pull request, so name who may edit it.
  • Each person's home files, such as ~/.claude/CLAUDE.md, belong to that person and should stay personal. A team rule that lives only there is a rule the rest of the team never gets (using one CLAUDE.md for all your projects).

What an owner cannot catch

An owner reviews what someone writes down. Most corrections never get that far: a teammate tells their agent “we retry on 5xx only” in a session, a reviewer says it in a pull request comment, the payments channel decides it in Slack. None of those opens a change to the rules file, so there is nothing to review. Getting them written down is its own job, covered in turning review comments into agent rules.

Where Harbor fits. Harbor gives the same ownership a place outside the files. Rules come in from agent sessions, pull request reviews, Slack and docs, and a capture setting decides which wait for a person and which go live on their own. Five roles in Spaces decide who reads a team's rules, who approves what goes in and who connects sources. Rules can carry an end date, and an opt-in timer archives rules agents have not used in 60, 90 or 180 days (knowledge and review). Rules imported from your existing CLAUDE.md and AGENTS.md wait for review before agents get them (rules files). The CODEOWNERS file above is still worth keeping for the files in the repo.

Questions

Who should own CLAUDE.md on a team?

Whoever answers for how the team's agents behave: one senior engineer on a small team, a group of two or three on a larger one. Name them in CODEOWNERS for CLAUDE.md, AGENTS.md, .claude/, .cursor/rules/ and the Copilot instructions, and turn on required review from code owners.

Can I use CODEOWNERS for CLAUDE.md?

Yes. A line such as "CLAUDE.md @acme/agent-rules" matches the file in every folder. Code owners are requested for review on any pull request that changes it, and "Require review from Code Owners" in branch protection makes their approval a merge condition. The last matching pattern in the file wins.

How often should we review CLAUDE.md?

On every change, through code owner review, and on a schedule for the rules that went stale without a change. Anthropic's docs ask for a periodic review to remove outdated or conflicting instructions, and /doctor prompt-audit finds references to files and commands that no longer exist.

Does code review keep agent rules up to date?

Only for what someone writes down. A correction given to an agent in a session, a pull request comment or a Slack decision never opens a change to the rules file, so the owner never sees it. Those need a way into the file first.

More on Team knowledge and shared conventions

  • Claude Code team memory: what is shared, what is not
  • Obsidian as a second brain for Claude Code, and where it stops
  • Share memory and rules across Claude Code, Codex and Cursor
  • How to stop AI coding agents repeating the same mistakes

How Harbor keeps a team’s knowledge →

Your team’s rules, in every agent, checked on every PR.

Harbor is shared memory for your team’s Claude Code, Cursor, Codex and Copilot. Free for you and one teammate. No card. No end date.

Start free →

Product

  • How it learns
  • Only what fits
  • Pull request review
  • Keep it true
  • Handoffs
  • The result
  • Security
  • For coding agents
  • Pricing

Developers

  • Docs
  • Quickstart
  • Install
  • CLI
  • MCP server
  • Served and cited
  • Rule-file checker
  • Blog

Guides

  • Why Claude ignores CLAUDE.md
  • AGENTS.md vs CLAUDE.md
  • Share Claude Code memory with a team
  • Why Codex ignores AGENTS.md
  • Why Cursor ignores rules
  • Copilot instructions vs AGENTS.md
  • Shared memory across agents
  • Copilot Memory and Dreaming

Compare

  • All comparisons
  • vs Mem0
  • vs Zep
  • vs Unblocked
  • vs ByteRover
  • vs Cognee
  • vs Hindsight
  • vs OpenViking
  • vs Supermemory
  • vs Glen
  • vs MemPalace
  • vs Sentra
  • vs Trail-code

Company

  • Questions
  • Contact
  • X
  • LinkedIn
  • Privacy Policy
  • Terms of Service
  • Data Processing Agreement
  • Refund Policy
© 2026 HarborConnect your agents