Product memory for Codex, Claude Code, and Cursor
RAAV keeps product truth, tasks, branches, decisions, verification, and audit in your repo so every coding agent starts from the same context.
Local-first. Zero-LLM by default.
Live sample workspace
Watch the founder loop play out — or click the sidebar to explore Today, Review, and Agents.
Founder control room
Decisions and proposals waiting for approval.
Managed agent identities with claimed work.
Open, in-progress, and blocked tasks.
Latest release and route checks.
This decides how safe multi-agent work feels for non-coding founders.
1.0 Understand
RAAV separates confirmed product facts, inferred assumptions, and open founder questions so Codex, Claude Code, and Cursor stop reinventing the roadmap.
Understand
Product
RAAV
Persistent product and project memory for agents building your product.
ICP
Non-coding solo founders and small teams coordinating Codex, Claude Code, Cursor, and MCP agents.
Promise
Every agent works from the same product truth, and every founder can see what is known, moving, blocked, and ready to review.
Needs founder review
Scope and promise include inferred assumptions separated from confirmed facts.
2.0 Prioritize
The Today view shows review count, active agents, open work, and verification — a founder control room instead of scattered chat history.
Prioritize
Decisions and proposals waiting for approval.
Managed agent identities with claimed work.
Open, in-progress, and blocked tasks.
Latest release and route checks.
This decides how safe multi-agent work feels for non-coding founders.
3.0 Direct
Every message includes the task, branch, claim, acceptance criteria, review policy, and submit command — not a blank chat window.
Direct
Show lanes, tickets, claims, risks, verification, product context, and audit trail by default.
codex/raav-public-launch-asset
apps/web/components/AuditExportViewer.jsx, apps/web/lib/raavDemoData.js
https://github.com/Yllasville/raav-ai/pull/214
4.0 Approve
Agents propose requirements, memory updates, and decisions. RAAV keeps them in review until a founder approves, defers, or rejects.
Approve
This decides how safe multi-agent work feels for non-coding founders.
1 proposal ready for approve / defer / reject
Add proposal review queue with confirm, defer, and reject before ledger writes land.
raav review proposals in your repo for the live queue from agent proposals.5.0 Verify
Lanes, claims, risks, verification results, and audit events stay attached to the work — not lost in terminal scrollback.
Verify
codex
Make the public launch asset prove install, memory, product, and audit flow.
claude_code
Polish Product Handbook language for non-coding founders.
cursor
Define Owner, Admin, Member, Viewer, and Agent roles.
pass
Next.js production build passed.
For the human founder
You do not manage raw prompts and scattered chat history. You manage the product memory, task queue, agent lanes, review decisions, and verification record.
Product Handbook separates confirmed facts, inferred assumptions, and open founder questions.
Goals, requirements, tickets, risks, and launch readiness stay connected to the work queue.
Agents read the same context, claim files, and work in visible branch/worktree lanes.
Proposal and memory review queues keep assumptions out of durable memory until you confirm them.
Each handoff records changed files, verification, risk notes, decisions, and an audit trail.
Non-coding founders are already building real software with frontier coding agents. The problem is not whether agents can write code. The problem is whether every agent knows the product, the current plan, the branch ownership, the open risks, and what the founder already decided.
RAAV is the shared memory and project management layer for that work.
Set up RAAV in this repository.
The founder pastes the install prompt into Codex, Claude Code, Cursor, or an MCP client. The agent initializes RAAV from inside the repo.
RAAV turns repo context, decisions, tasks, and founder input into a Product Handbook with confirmed facts, inferred assumptions, and review gaps.
Agents read the same overview, claim tasks and files, work in branch/worktree lanes, and submit verification back into the audit trail.
RAAV shows the founder what is known, what is moving, what is blocked, and what is ready to review.
Product Handbook, ICP, promise, V1 scope, non-goals, and founder review gaps.
Tasks, tickets, owners, branch lanes, active claims, and file conflicts.
Verification status, risk notes, decisions, audit chain, and next action.
| Without RAAV | With RAAV | |
|---|---|---|
| Product memory | "Let me explain the app again..." | raav product handbook shows confirmed facts, assumptions, and open founder questions |
| Agent coordination | Two agents can edit the same files blindly | Tasks, branch/worktree lanes, claims, and conflict checks |
| Founder visibility | Raw terminal logs and scattered chat history | Founder-readable product brief, next decision, risks, and audit trail |
| Work handoff | Nobody can reconstruct what changed or why | raav submit records files, summary, verification, risks, and next action |
| AI spend | Another product tries to reason with its own model | Zero-LLM by default; Codex and Claude Code do the reasoning |
I am Ville Ylläsjärvi. I built FeedSquad with AI agents from inside the Arctic Circle, solo.
RAAV exists because the hard part is no longer asking an agent to code. The hard part is keeping product intent, tasks, branches, decisions, and handoffs coherent across many agent runs.
Free locally. Pay for hosted visibility, coordination, and audit — not agents or tokens. Prices exclude taxes.
Full local memory for one repo
Hosted console for one founder with 3 projects · €10/mo billed yearly
2-seat minimum · agents are never billed as seats