AGENTS.md template

An AGENTS.md template that becomes reviewable memory.

Start from an AGENTS.md template, then upgrade it into memory you can review: RAAV syncs agent rules and adds a review queue, swimlanes, and audit trail.

What RAAV gives agents

  • Keep an AGENTS.md your agents read, generated and synced by RAAV
  • Upgrade static rules into memory the founder can review
  • Keep live work state outside the static instruction file
  • Swimlanes and audit make the collaboration visible

The problem

An AGENTS.md template is a good start, but a flat rules file cannot separate confirmed facts from assumptions, show who is editing what, or record verification.

The RAAV layer

RAAV keeps AGENTS.md as the short instruction sheet it should be, then backs it with current project state. Agents get rules plus scoped work; founders get proposals, ownership, and run history.

A template is a starting point, not the whole memory

An AGENTS.md or CLAUDE.md gives the agent baseline instructions. RAAV keeps that file and adds the changing project state a founder needs to review.

What belongs in an AGENTS.md

A useful AGENTS.md covers project setup, coding conventions, test commands, and the rules for claiming files and submitting verification.

RAAV generates and syncs these rules across AGENTS.md, CLAUDE.md, Cursor rules, and Copilot instructions so every agent follows the same protocol.

  • Setup, conventions, and test commands
  • Operating rules: read, claim, check, submit
  • Synced across agent rule files

Separate repository rules from product decisions

A flat AGENTS.md changes silently and hides assumptions. RAAV separates confirmed facts from inferred ones and routes changes through a review queue.

Stable instructions can stay in AGENTS.md while product changes wait in a founder review queue.

  • Confirmed vs inferred facts
  • Proposals instead of silent edits
  • Founder sign-off before durable writes

AGENTS.md cannot show who is working now

Swimlanes and claims show which agent is applying the rules now. Submitted work records what changed and which checks passed.

The template is where you start; reviewable memory is where you end up.

  • Swimlanes for active agent work
  • Verification recorded on submit
  • Audit export for founder review

How the agent loop works

RAAV does not replace your coding agent. It gives the agent a durable operating system for product truth, task ownership, and proof.

Generate

RAAV writes an AGENTS.md and matching rules for your agents.

Sync

The same operating rules land in CLAUDE.md, Cursor, and Copilot files.

Operate

Agents read memory, claim files, and submit verification.

Review

The founder supervises via the review queue and audit trail.

Generate an AGENTS.md that is more than a file

Ask your agent to install RAAV and generate the AGENTS.md for you. You get synced rules plus a review queue, swimlanes, and audit.

Paste this into your coding agent
Install RAAV in this repo and generate an AGENTS.md template. Sync the operating rules across CLAUDE.md and Cursor, and make agents claim files and submit verification as part of those rules.

AGENTS.md remains the instruction file. RAAV records the changing work and decisions around it.

Why this is different

Most tools either write code or track human tickets. RAAV sits between the founder and the coding agents as shared product memory, coordination, and audit.

Generates and syncs AGENTS.md, CLAUDE.md, Cursor, and Copilot rules
Instruction syncing and project records do not require RAAV to call an LLM
The €10/week Trial includes local instructions and the hosted console once invited

A flat AGENTS.md vs AGENTS.md backed by reviewable memory

Change control
The file changes silently; the founder never sees it.
Changes route through a review queue.
Watch
No view of which agent is acting on the rules.
Swimlanes and claims show active work.
Supervise
No record of what shipped or passed.
Audit trail records verification and risk.

FAQ

Founders ask whether they still need an AGENTS.md if they use RAAV.

Do I still need an AGENTS.md?

Yes, and RAAV generates and syncs it. The file is real memory your agents read. RAAV adds the review queue, swimlanes, and audit that a flat file cannot provide.

Does RAAV overwrite my existing rules?

RAAV syncs operating rules across agent files while keeping your project-specific content. The point is one consistent protocol across agents.

Is RAAV a coding agent?

No. RAAV prepares and maintains the context files that Codex, Claude Code, Cursor, and Copilot read.

What does it cost?

The Trial is €10/week. The plan includes the local instruction workflow and a hosted founder console; invitations are still released in small cohorts.

AGENTS.md template for product memory | RAAV