Pricing

Free locally. Paid where intelligence compounds.

Tavsin is free to run on your machine — that's the local-first promise. The paid tiers exist for one reason: a team's accumulated intelligence is worth paying to keep, share, and protect.

Free

Local

Everything you need to ship from one machine.

$0 / forever
No card. No telemetry. No expiry.
  • Desktop app + CLI
  • Software Intelligence Graph
  • Local memory store (you own it)
  • Outcome contracts & trace history
  • Bring your own model (Anthropic / OpenAI / Gemini / local)
  • Autonomous agent mode
  • MCP server access
  • Team memory sync
Download

No account required.

Most popular

Pro

For individuals shipping serious work.

$20 / month
Billed monthly · cancel any time
  • Everything in Local
  • Autonomous Agent mode + boundaries
  • MCP server — plug Tavsin into Claude Desktop, Cursor, Zed
  • Unlimited trace history & project size
  • Cloud sync of memory across your machines
  • Priority support
  • Shared team memory
Start free trial

14-day trial · no card up front.

Compounding

Team

A shared memory layer your team builds together.

$40 / seat / month
Min 3 seats · billed annually or monthly
  • Everything in Pro, per seat
  • Shared compounding memory — one teammate's fix becomes everyone's rule
  • Org-wide intelligence graph
  • SSO (Google / Microsoft / GitHub)
  • Audit log & role-based access
  • Centralised provider keys & budget
  • Onboarding session for the team
Start Team

Or self-host for free, paid features locked.

Enterprise

Enterprise

For organizations with regulatory or scale needs.

Custom
Annual contract
  • Everything in Team
  • On-premises deployment (VPC / air-gapped)
  • Bring your own LLM endpoint
  • Dedicated success engineer
  • Custom integrations & training
  • 99.9% SLA, security review, DPA
Talk to us

SOC2 in progress · GDPR-ready.

Why this model

We don't charge for tokens. We charge for what stays.

Per-token markup is what every other AI-coding tool has settled on. We don't, on principle. The model is interchangeable; the intelligence you build on top of it isn't.

Free Tavsin already pays for itself: you avoid the markup, you keep your provider relationship, and your project memory lives on your disk. The paid tiers exist for the things only money makes possible — running agents in the cloud while you're away, syncing memory across machines, and the one feature no individual workflow can build alone: a team's shared, compounding intelligence layer.

The math

$0.00

Tavsin's markup on your model usage.

100%

Of your intelligence stays on your machine — even on the paid tiers, by default.

~3×

Typical multiplier on the value of a fix once it lives in your team's memory.

Questions worth asking

Is the free tier real?
Yes. The desktop app, CLI, graph, memory, outcomes, and traces are free forever, locally. No "free trial that expires" trick — the local version is the product.
What does Pro actually unlock?
Autonomous Agent mode (the agent runs end-to-end with optional strict boundaries), the MCP server (so Claude Desktop / Cursor / Zed can call Tavsin's tools), unlimited trace history, and cloud sync of memory across your machines.
What's the Team tier's killer feature?
Shared compounding memory. When one teammate's fix is approved, it becomes a rule the whole team's agents respect. No other AI tool does this — because no other tool has a memory layer to share.
Do I have to use your cloud?
No. Self-host the workspace anywhere (your laptop, a VM, on-prem). Pro/Team features unlock with a license key; the data stays where you put it.
Can I bring my own model?
Always. Anthropic, OpenAI, Gemini, or any OpenAI-compatible local endpoint. You pay them directly; we never see your keys.
What about open source?
The engine and free-tier features are open source. Paid features are closed but verifiable — the gates live in clear license-key checks.