Pricing
The platform is free. The only bill is the model, paid with your own keys, credits at 0% markup, or your users' own subscriptions. Contracts are for enterprises that need guarantees.
The full platform, not a trial. The only bill is the model's own.
- Included1 agent
- IncludedHosted docs
- IncludedProvenance engine
- IncludedUS-based data locality
- IncludedGenerous rate limits
- IncludedCommunity support
Everything in Free, plus the guarantees enterprises need.
- IncludedUnlimited agents
- IncludedDocs on custom domain
- IncludedDedicated rate limits
- IncludedSAML single sign-on
- IncludedUptime SLA
- IncludedFDE + Deployment Strategist
- IncludedDedicated Slack channel
Platform
Security & support
Available models
The only bill is the model's own, at cost with no platform markup. Here is exactly what every model your agents can run on costs, and none of them train on your data.
| Model | Trains on data | Input | Output | Cache read | Cache write 5m | Per request |
|---|---|---|---|---|---|---|
| Claude Haiku 5.5 | No | $0.10 | $0.50 | $0.01 | $0.13 | $0.0024 |
| Claude Sonnet 5.5 | No | $2.00 | $10.00 | $0.10 | $2.50 | $0.0440 |
| Claude Sonnet 5 | No | $2.00 | $10.00 | $0.20 | $2.50 | $0.0480 |
| Claude Sonnet 4.6 | No | $3.00 | $15.00 | $0.30 | $3.75 | $0.0720 |
| Claude Opus 5.5 | No | $4.00 | $20.00 | $0.20 | $5.00 | $0.0880 |
| Claude Opus 5 | No | $5.00 | $25.00 | $0.50 | $6.25 | $0.1200 |
| Claude Opus 4.8 | No | $5.00 | $25.00 | $0.50 | $6.25 | $0.1200 |
| Claude Opus 4.7 | No | $5.00 | $25.00 | $0.50 | $6.25 | $0.1200 |
| Claude Opus 4.6 | No | $5.00 | $25.00 | $0.50 | $6.25 | $0.1200 |
| Claude Fable 5.1 | No | $10.00 | $50.00 | $0.25 | $12.50 | $0.2100 |
| Claude Fable 5 | No | $10.00 | $50.00 | $1.00 | $12.50 | $0.2400 |
Rates in USD per million tokens, billed at cost as your agent runs and itemized to the conversation that caused it.
FAQ
No. The full platform is free, not a trial. The only money in the system is the model's own bill, and none of it is ours.
You add an API key from Anthropic or OpenAI and your agents call the model with it. The usage lands on the provider bill you already have, at whatever rates you've already negotiated, and never passes through us. Swap or remove the key whenever you like.
You buy credits and we pay the model's bill on your behalf, at exactly its list price. 0% markup, no fees on top. Every call lands on the conversation that caused it, so the bill is auditable line by line.
Your users can connect the ChatGPT subscription they already pay for, and their own conversations run on it. It covers that user and nothing else, never another user's conversations and never your platform usage. For everyone who opts in, their questions cost you nothing to answer.
They aren't exclusive. Turn on all three and they settle in order. A user who has connected their own subscription runs on it; everything else draws on your API key for the providers it covers, and your credit balance covers whatever the key can't. Start on credits if you'd rather not manage a provider key, add a key if you already have provider rates worth using, and switch on end-user subscriptions when your users are the ones asking the questions.
No. Start on your own model key, or add a card when you top up credits. Either way the platform itself stays free.
Work parks. Nothing fails and nothing is lost: the moment funds land, everything that was waiting resumes exactly where it stopped.
About a day: connect the repo, review what the agent learned, approve the actions you want it to have, and drop one script tag into your product.