Pay when you grow.
Not before.
Start with a useful cloud workspace, not a demo. Move to Pro when you need more active projects and context capacity. Use Enterprise when the deployment boundary itself is custom.
A real cloud workspace for solo projects. No card required.
Start free- 1 active cloud project
- 50K context credits every 30 days
- 1.0 GB retained context
- Core map, memory, and MCP reads stay available
- Local mode remains free and unlimited
Serious context infrastructure for solo builders and small teams.
Start Pro- Unlimited active projects
- 500K context credits per billing period
- 10.0 GB retained context
- Hosted remote MCP: a project-scoped endpoint any MCP tool can read
- Continuous auto-freshness: CI keeps your cloud map current on every push
- $5 adds 250K non-expiring credits
- AI memory assist: auto-distilled proposals from your code
- Schema-drift distillation: DB schema changes become reviewed memory proposals
- Issue and PR ingest: harvest decisions from GitHub discussions into memory
- Validated memory + history, diff and rollback
- Living knowledge base: 11 auto-written, self-updating docs (architecture, API, data model, decisions…)
For organizations with their own trust boundary and deployment requirements.
Talk to us- Dedicated US, EU, or TR deployment options
- Self-hosted or isolated infrastructure
- SSO / SAML, audit logs, and policy controls
- Custom storage, context, and support terms
The “reads ~5× less” isn't a slogan — it's measured on five real repos. See the benchmark →
What a credit buys
Credits meter the paid AI work, so each plan's number means something concrete. Reading the map, retrieval, and reading memory cost nothing. Only the generated AI operations draw down credits, at these rates:
- An answer from /ask1 credit
- A distilled memory proposal1 credit
- A memory conflict check1 credit
- A generated living doc1 credit
- An embedding0.1 credits
So Free's 50K credits and Pro's 500K are roughly that many answers or distills per period, or ten times as many embeddings.
What each plan actually enforces
No surprises: credits are metered as you use them, and usage is capped when you pass your plan allowance. Beyond credits, each plan carries a couple of live monthly limits so spend stays predictable.
Local mode stays unlimited and free, on any plan. Hit a limit and nothing breaks: you top up or upgrade, and existing work keeps syncing.
Questions, answered
Is the free tier actually free?
Yes. Local mode stays free and unlimited. The cloud Free plan needs no card and includes 1 active project, 50K context credits every 30 days, and 1.0 GB retained context.
What leaves my machine?
The map (symbols, signatures, routes) and embeddings. Never your source, never your secrets. On Enterprise, nothing leaves at all.
Do you charge per query?
No per-query invoice. Pro is $10 per workspace and includes 500K context credits each billing period. If you need more, a $5 top-up adds 250K credits that never expire.
Can I self-host?
Yes, on Enterprise. Run the whole map + memory layer inside your own trust boundary so nothing crosses your perimeter.
Which AI tools does it work with?
Any tool that speaks MCP: Cursor, Claude Code, GitHub Copilot, Cline, Windsurf, Zed, Continue, Aider. One sync, they all read it.
Can I export my memory?
Always. The map and memory are yours — export them, take them with you. No lock-in is part of the point.
Start on the free tier. Really free.
> daemon: $0 free tier. actually free. no credit card.