Pricing that scales with your agents
Start free. Pay only when your agents go to work. No seat licenses, no hidden inference markup.
Starter
For solo builders shipping their first agent.
BILLED MONTHLY
Start free trial- 1 workspace, 3 agents
- 50k tokens / month included
- nexus-1-turbo model
- Community support
- 7-day run history
Pro
For teams running agents in production.
BILLED MONTHLY
Start free trial- 5 workspaces, unlimited agents
- 1M tokens / month included
- All models, including nexus-1-pro
- Priority support (4h SLA)
- 90-day run history + traces
- Custom tools & webhooks
Enterprise
For orgs with compliance and scale needs.
BILLED MONTHLY
Talk to sales- Unlimited workspaces
- Volume token pricing
- Dedicated capacity + private region
- 24/7 support, 30-min SLA
- SSO/SAML, audit logs, SOC 2 report
- Named solutions engineer
Compare every plan
Prices shown for monthly billing.
| Feature | Starter$29/MO | Pro$99/MO | Enterprise$499/MO |
|---|---|---|---|
| Included tokens / month | 50k | 1M | Custom |
| Requests per minute | 60 | 600 | Custom |
| Concurrent runs | 3 | 25 | Unlimited |
| Streaming API | |||
| Custom tools & webhooks | |||
| Run traces & replay | |||
| SSO / SAML | |||
| Audit logs | |||
| Private region | |||
| Support | Community | 4h SLA | 30-min SLA |
Frequently asked questions
Roughly four characters of English text. Both the prompt you send and the completion the model returns are billed. The Playground shows the exact count for every run, so you can size your bill before writing a line of code.
Nothing breaks. Overage is billed at $0.60 per 100k tokens on Starter and $0.40 on Pro. You can also set a hard cap in the workspace settings and the API will return a 429 instead of charging you.
Yes. Upgrades take effect immediately and we prorate the difference. Downgrades apply at the end of the current billing period, so you keep the higher limits until then.
Every new workspace starts with 5,000 free tokens and full access to nexus-1-pro for 14 days. No credit card required to sign up.
Never. Prompts and completions are encrypted at rest, retained only for the run-history window of your plan, and are never used to train models. Enterprise workspaces can set retention to zero.
Still deciding?
Try the models for free