Pricing
Free to start. Paid plans add more projects and concurrent agents, BYOK, bundled frontier allowance, Slack routing, and team workflow — with inference billed at provider cost, never marked up.
Start free
Try it on one repo, free.
Start free on one repo.
Ship every day
One repo, every day. Bring your own model keys, or run on bundled inference.
You already have API keys. Pay platform, not inference.
One repo, shipped every day.
Run a portfolio
Every repo, in parallel — for operators and small teams.
Portfolio operator running several products.
Tiny team or small agency.
Every plan includes
What a task costs
A typical small task costs cents to a few dollars of inference — you only pay for what the agent actually uses, at provider cost.
Set a per-task spend limit. If a task hits it we stop the task and refund the platform fee, so the cost of every task stays predictable. Retries on the same task are free.
Recurring workflows (/loop) and your own guardrail hooks are included on every plan — no separate add-on. A loop is just goals on a schedule: each iteration respects the same per-task cap, and you set a cap on the whole loop too, so an unattended fleet stays predictable.
$example: a "fix the flaky test" task ran 3 iterations and cost $0.40 of inference.
Billing
Bring your own key. On any paid plan (from the $10/mo BYOK plan up), drop in an Anthropic, OpenAI, Google, OpenRouter, or custom OpenAI-compatible API key and the agent runs on it — no markup. You pay the platform fee; inference cost is what you'd pay anyway. Free runs on the included GLM 5.2 + Kimi K2.7 Code allowance.
A task is one plain-English request that ends in a PR (or a failure log). Retries on the same task are free. A task that runs three iterations to land a PR still counts as one.
No. Every plan is usage-based — you pay for the inference you actually consume, billed at provider cost. Free and BYOK include $100 of GLM 5.2 + Kimi K2.7 Code free, one-time; Solo/Fleet/Pro include a recurring $100/mo of GLM 5.2 + Kimi plus a bundled frontier allowance ($20/$80/$200 of Claude, GPT, and Gemini). Past your allowance you upgrade for more headroom, or bring your own key.
You get an email as you approach it. On Free and BYOK the $100 of GLM 5.2 + Kimi is a one-time allowance — agents pause once it's used up until you upgrade. On Solo/Fleet/Pro the $100/mo of GLM 5.2 + Kimi plus the bundled frontier allowance reset each billing cycle. On a paid plan, bring your own key to keep running on your own inference, billed by your provider.
Self-serve through the billing portal. Service continues to the end of the current billing period. Projects, agents, and task history archive for 30 days, then delete.
7-day money-back guarantee on first month, no questions asked. After that, prorated to the next billing cycle.
Monthly only at launch. Annual (with two months free) ships once we have 90+ days of churn data to set the discount sensibly.
Get started
Connect a GitHub repo, hand off a task in plain English, and get a reviewed Pull Request back. Start free and upgrade when your fleet needs more room.
free to start · usage billed at cost · cancel anytime