Simple plans. No surprise bills.
Every plan is a flat monthly fee with an included allowance of agent usage and sandbox compute. Reach your limit and agents simply pause — you're only billed beyond it if you opt in.
Kick the tires on a real repo.
- The complete platform
- No credit card required
For developers shipping every day.
- Everything in Free
- Pay-as-you-go available
- Optional monthly spend cap
For heavy, daily agent work.
- Everything in Pro
- Triple the monthly allowance
Maximum throughput, no sandbox limits.
- Everything in Pro+
- Unlimited sandbox compute
All prices in USD, billed monthly. Included usage resets every cycle.
Pay for the work, never for surprises
Usage is metered in dollars of model work, with hard stops you control. Here's exactly how it adds up.
Usage measured in dollars
Your plan includes a monthly allowance of agent usage — the list-price cost of the model work your agents do as they plan, write code, and run tools. Watch it draw down live in the app.
No surprise bills
Pay-as-you-go is off by default. When you reach your included allowance, agents simply pause until your next cycle. You're never charged a cent you didn't opt into.
Pay-as-you-go, on your terms
On a paid plan, switch on pay-as-you-go to keep working past your allowance — billed at cost. Set an optional monthly spend cap and usage stops the moment you hit the number you chose.
Sandbox compute, included
Every plan includes a monthly budget of sandbox hours for builds, tests, dev servers, and driving the real UI. Idle and paused sandboxes don't count, and there's no per-hour charge.
Change or cancel anytime
Upgrade instantly — you pay a prorated amount for the rest of the current period and your larger allowance is available right away, with the same renewal date. Downgrades and cancellations take effect at the end of your billing period; you keep full access until then.
What's in every tier
Every plan gets the complete platform. Tiers differ only by how much agent work and sandbox compute they include.
| Plan | Free | Pro | Pro+ | Ultra |
|---|---|---|---|---|
| Monthly price | $0 | $20 | $60 | $200 |
| Included agent usage | $10 / mo | $20 / mo | $60 / mo | $200 / mo |
| Sandbox compute | 2 hrs / mo | 20 hrs / mo | 60 hrs / mo | Unlimited |
| Pay-as-you-go overage | — | |||
| Optional monthly spend cap | — | |||
| Full platform — every feature |
What your allowance buys
Agent usage is metered at provider list prices — the same per-token rates you'd pay the model maker directly, with no markup. Your monthly allowance is drawn down against these.
| Model | Input | Output | Cached input |
|---|---|---|---|
| Claude Fable 5Anthropic · 1M context | $10.00 | $50.00 | $1.00 |
| Claude Opus 4.8Anthropic · 1M context | $5.00 | $25.00 | $0.50 |
| Claude Sonnet 5Anthropic · 1M context | $2.00 | $10.00 | $0.20 |
| GLM-5.2Z.ai via OpenRouter · 1M context | $2.00 | $6.00 | $0.20 |
| Grok 4.5xAI via OpenRouter · 500K context | $2.00 | $6.00 | $0.50 |
| GPT-5.6 SolOpenAI · 1M context | $5.00 | $30.00 | $0.50 |
| GPT-5.6 TerraOpenAI · 1M context | $2.50 | $15.00 | $0.25 |
| GPT-5.6 LunaOpenAI · 1M context | $0.75 | $4.50 | $0.075 |
Prices are in USD per 1M tokens. Anthropic models, GLM, and Grok bill at a flat rate across their full context window. Sonnet 5 is on launch promo pricing ($2 / $10) through August 2026. GPT-5.6 Sol and Terra requests above 272K tokens bill at a higher long-context rate (input and cache ×2, output ×1.5).
Questions, answered
What is “agent usage” measured in?
Dollars of model work — the list-price cost of the tokens your agents consume as they plan, write code, and run tools. Every plan includes a monthly dollar allowance, and you can watch it draw down live in the app.
What happens when I reach my included allowance?
By default your agents pause until your next monthly cycle — no surprise charges. Upgrade to a higher plan, or switch on pay-as-you-go to keep working.
How does pay-as-you-go work?
It's off by default. On a paid plan you can turn it on to continue past your included allowance, billed at cost. Set an optional monthly spend cap and usage stops the moment you reach the number you chose.
What are sandbox hours?
Each thread runs in its own isolated cloud sandbox. Sandbox hours meter the real time those sandboxes spend running — builds, tests, dev servers, and driving the UI like a human. Each plan includes a monthly budget; idle and paused sandboxes don't count.
Do I pay for sandbox time on top of my plan?
No. Sandbox hours are included in your plan as a monthly budget, with no per-hour overage.
Can I change or cancel my plan?
Anytime. Upgrades apply immediately — you pay a prorated amount for the rest of the current billing period and your larger allowance is available right away; your renewal date doesn't change. Downgrades and cancellations take effect at the end of your current billing period, so you keep your plan and allowance until then.
What does it cost to upgrade mid-cycle?
A prorated share of the price difference, based on the time left in your billing period — you get credit for the unused time on your old plan and pay the remaining days at the new plan's rate, charged immediately. Halfway through the month, upgrading from Pro ($20) to Pro+ ($60) costs about $20. Downgrades are never charged; they simply take effect at your next renewal.
When does my included usage reset?
At the start of each billing period. On a paid plan the period follows your subscription date; on Free it's the calendar month. Changing plans doesn't reset the period — an upgrade raises your allowance for the rest of the current period, and usage you've already spent counts against the new, larger allowance.
Is my code private?
Yes. Sandboxes are isolated per thread and torn down after use. Private repos authenticate with a token you control.
Where do agents run?
Each thread runs in its own isolated cloud sandbox, branched from your repo. Heavy builds and test suites never touch your laptop.
Ship your next feature with a swarm
Spin up your first project in under a minute. Bring a git repo or start from an empty one.