Pricing
Free until we
can prove it's worth paying for.
We're early. Pro is coming, but until it lands, every team gets full access — we eat the costs. That's not a marketing line; it's how we'd want to be sold to.
Build, test, and ship one production agent.
Join the waitlist →- ✓ 1 production agent
- ✓ 1 project (dev + prod deployments)
- ✓ Platform + SDK access
- ✓ 10K agent runs / month
- ✓ 100K LLM gateway tokens / month
- ✓ 1 GB runtime state (memory + indexes)
- ✓ Community support
Pro for everyone we let in. We're admitting teams in batches so we can actually help you ship.
Join the waitlist →- ✓ 3 production agents
- ✓ 1 project · dev + staging + prod
- ✓ 5 team seats
- ✓ 100K agent runs / month
- ✓ 2M LLM gateway tokens / month
- ✓ Eval suites + regression scores
- ✓ Replay + 90-day audit log
- ✓ Approval flows
- ✓ Encrypted secret vaults + BYO model keys
- ✓ Priority Discord + email
You've outgrown Pro — or you need security, residency, and volume Pro can't deliver.
Talk to us →- ✓ Unlimited agents, projects, runs
- ✓ Unlimited team seats
- ✓ Indefinite audit log retention
- ✓ Self-host or BYOC deployment
- ✓ SOC 2 Type II + DPA
- ✓ Custom data residency
- ✓ SLA + dedicated support
- ✓ SSO / SCIM
- ✓ Volume discounts on usage
Not sure if we fit?
Walk through your use case with us before signing up. 20 minutes, no pitch deck.
When we turn on billing
What we'll meter.
Three usage dimensions on top of the per-seat price. Predictable, capped, and observable in the app.
In + out across every model call — agent reasoning AND the stateless /chat/completions surface. One Tavora bill. BYO keys (via secret vault) lights up later and skips this dimension entirely.
e.g. 4M tokens / mo · ~$8 at GPT-4o-mini rates
A run is one user-initiated turn — including all tool calls and code execution within it. Counted across the project.
e.g. 50K runs / mo across the project
Memory stores + RAG indexes Tavora holds for the agent. Your business data stays in your backend (PocketBase / Supabase / your DB) — Tavora reads it via MCP, never copies it.
e.g. 12 GB of agent working memory + RAG indexes in your project
FAQ
Things you'd reasonably want to know.
When we've seen enough production usage to price it confidently. Probably 6–9 months out. You'll get 60 days notice and the published rate is what you'll pay — no surprise tiers.
Yes — on Pro and Enterprise. Put your provider key in a Tavora secret vault (envelope encryption, never returned over the API), designate that vault as the project's LLM source, and every agent run / chat-completion call uses your key instead of ours. Mix providers freely: vault holds your OpenAI key, leave Anthropic on the Tavora gateway, fall back to global for anything you didn't override. We never see the keys — the runtime reads them inside the sandbox, encrypted at rest, never logged.
One user-initiated turn — from the moment a user sends input to the moment the agent stops. All tool calls, model calls, and sandboxed code execution within that turn are part of the same run.
Yes. Free stays free, with the limits listed. It's for solo developers, side projects, and teams kicking the tires. Pro is for teams shipping to customers.
When you hit Pro's ceiling on agents (3), seats (5), runs (100K/mo), or data (25 GB) — or when you need multi-project, SSO, a signed DPA, self-host, custom residency, indefinite audit retention, or an SLA. Either email us before you hit the wall, or we'll reach out when you do.
Available on Enterprise. Same binary, same app, runs in your VPC. We can also do dedicated single-customer cloud if you want managed but isolated.
No catch. We need teams shipping real workloads so we can build the right runtime. Free Pro is how we trade your usage data (which we can see in our infra) for our runtime (which you can use). Fair trade.
Get on the list. Build something. Ship it.
Pro is free for everyone we let in. Join the waitlist for the next batch, or talk to us if you want to skip the queue.