Build and run locally, free, forever (open source). Subscribe to run hosted.
Hosted runs for one person, just you, automating your own work.
For a team running shared workflows. Every member is billed, and allowances pool across the org.
Org-wide and procurement-grade, with committed usage and dedicated capacity.
Solo and Pro are single-member plans: add anyone else and the org moves to Team, where every member is $40 per month regardless of role (no free seats). You can choose Team at any size, even with one member. Allowances reset monthly with your billing cycle and never roll over.
Every paid plan includes monthly agent-hours: one agent-hour is an hour on a standard 1-vCPU machine, and larger machines draw the allowance proportionally. Beyond the allowance, compute bills per active minute at these rates. When a run suspends, on a long sleep, a human-input gate, or while a child run does the work, its machine is released and that time is free.
| Machine size | Shape | Rate |
|---|---|---|
| Small | 1 vCPU · 2 GB | $0.05 / min |
| Medium | 2 vCPU · 4 GB | $0.10 / min |
| Large | 4 vCPU · 8 GB | $0.20 / min |
| XL | 8 vCPU · 16 GB | $0.40 / min |
On the managed lane, the model's rate at cost, with no per-token markup. The pool equals your plan price, so the whole subscription is convertible to model usage ($20 on Solo, $100 on Pro, $40 per member on Team), and it works on any surface, including local dev. Bring your own key and tokens are $0 from us.
See all models and rates →| Model | Input / Mtok | Output / Mtok |
|---|---|---|
| GPT-5.5 | $5.00 | $30.00 |
| Gemini 3 Pro | $2.00 | $12.00 |
| Claude Opus 4.8 | $5.00 | $25.00 |
Let a run search the live web, no key of your own required. Plans include 500 searches a month on Solo, 2,000 on Pro, and 1,000 per member on Team; beyond that, $0.008 per search, and a deeper search counts as two.
Run outputs and persisted codebases (opt-in workspaces) are stored for you. Free includes 5 GB as a hard cap, so it never bills for storage. Paid plans include their allotment, then $0.25 per GB-month from prepaid credits, prorated so deleting it stops the charge. Run outputs auto-expire after 90 days.
| Plan | Included | Overage |
|---|---|---|
| Free | 5 GB | Hard cap |
| Solo | 20 GB | $0.25 / GB-month |
| Pro | 100 GB | $0.25 / GB-month |
| Team | 100 GB + 25 GB / member | $0.25 / GB-month |
Everything you need to build and run locally, forever. Free covers up to 3 members building workflows, deploying drafts, and running them on their own machines with the open-source engine and their own model key. Hosted runs, managed inference, and self-hosted runners start on a paid plan. You can deploy cron and webhook triggers on Free, but they stay paused until you subscribe.
They are individual plans: just you, automating your own work. The moment you add anyone else, the org moves to Team, where every member is $40 per month regardless of role and allowances pool org-wide. You can also choose Team at any size, even with one or two members, if you want pooled allowances and its governance controls.
One hour of active run time on a standard 1-vCPU machine. Larger machines draw the allowance proportionally, so an hour on a 4-vCPU machine uses 4 agent-hours, matching the per-vCPU compute rate. Only active execution counts: when a run suspends on a long sleep, at a human-input gate, or while a child run does the work, its machine is released and that time is free.
Every paid plan includes monthly allowances: agent-hours, a managed-token pool at cost, web searches, and storage. The dashboard shows what's left in real units, like 14.2 of 25 agent-hours. When an allowance runs out, usage keeps working as overage drawn from prepaid credits at the public rates on this page, identical on every plan: compute at $0.05 per vCPU-minute, managed tokens at cost, storage at $0.25 per GB-month, and web search at $0.008 per search. An org spend cap keeps overage inside a ceiling you set.
No. Allowances reset in full on your monthly billing anniversary, and unused amounts don't carry forward. Prepaid credits are the opposite: whatever balance you buy stays until you spend it and never expires.
Usage itself carries no markup: managed tokens bill at the model's source price, and compute, storage, and search bill at the public rates. The only fee is 10% when you top up prepaid credits, so a $50 top-up costs $55 and adds $50 of credit. It covers card processing and the upstream credit fee we pay to route your tokens, which is what keeps the per-model rates at cost.
They pause, and they pause loudly. On Free, triggers deploy but don't fire. If a subscription ends or an org downgrades, its triggers pause the same way and you're notified in the dashboard and by email; nothing is deleted. When you resubscribe, resuming them is one click.
Run outputs (artifacts) and opt-in persisted workspaces (codebases that carry across runs) are stored for you. Free includes 5 GB as a hard cap and never bills for storage. Solo includes 20 GB, Pro 100 GB, and Team 100 GB plus 25 GB per member. Beyond the included amount, storage is $0.25 per GB-month drawn from the same prepaid credits, prorated so deleting storage stops the charge. Run outputs auto-expire after 90 days by default. If an account stays unpaid past a grace period, stored data may be removed.
Two ways. The SDK, CLI, and single-node runtime are open source, so you can run everything yourself with no account and no plan, free forever, and the same workflow file moves between local and hosted unchanged. Paid plans also include self-hosted runners: your machines attached to the hosted control plane, 1 on Solo, 3 on Pro, and 10 per org on Team. Compute on your own runners is never billed; brokered usage like managed tokens, storage, and web search bills normally.
Your first workflow runs on your own machine today, free. A plan makes it hosted: durable, triggered, and always on.
Prices in USD and may change before general availability.
