Your own private AI coding workspace

Add-ons

Two things you can buy on top of any plan:

• Data transfer — how much data your workspace may send out each month: files you download, uploads to the people using your app, and API responses. (Hosts call this “egress”; incoming data is always free.) Everyday coding and chat fit inside the included allowance — you only need more for heavy media work or moving large datasets.
• Disk space — the room your files, repositories and build artifacts take up on your virtual machine. Start small and grow it any time; it can grow but not shrink, so you are never charged for space you gave back.

How to add these: pick them when you subscribe, or change them at any time afterwards in Settings → Billing — the change is prorated, and your workspace keeps its files. Data transfer is metered on your own virtual machine, which warns you before you reach the allowance. How add-ons work, in detail.

Fleet slots — run more than one coding agent at once

Your instance is the captain. A slot is a second virtual machine that runs a coding-agent CLI of your choice — Kimi CLI, Qwen Code CLI, or our own harness — on your repositories, working in parallel with you and with each other. Every task gets its own isolated copy of the repo, so nothing touches your working tree until you decide to merge it.

• Parallel delivery. On our pilot's 30-task file suite the fleet merged 30/30 first-pass at 1.97× the wall-clock speed of a single harness — for about $0.0004 of model spend.
• A second opinion, built in. Send the same job to two different harnesses and compare, or have one review another's diff before it ever reaches your branch.
• The right agent for the job. Our public sweeps of the same 6-task suite: Kimi CLI 32.5s, Qwen Code CLI 85.5s, our own harness 119.9s. Keep the fast one on the small jobs and the patient one on the hard ones.
• Nothing stops the work. If a slot runs out of memory it gets one free grace run on the next size up, so the job still completes. Three memory limits in a month and you get a one-click upgrade — you are never blocked and never force-billed.
• Your keys, your quotas. Slots are BYOK: tasks run on your own provider keys, so your plan, rate limits and discounts apply. We do not mark up tokens.

How to get a slot: buy it yourself from your workspace — pick a size in the Fleet tab and add it to your plan. It is billed with your plan, per slot per month, from the moment you add it, and you can cancel any slot from the same place. Nothing is charged until you confirm. A cancelled slot stops immediately, is charged through the end of the month it is in, and is not refunded. How slots are added.

Per slot, per month — cancel any time. Launch pricing is a month-to-month discount: we may withdraw it at any time, and the standard price then applies from your next billing period. Rates are not locked in and no rate is grandfathered. Add slots to any plan; each one is its own virtual machine, so one busy slot never slows down another.

Have a promotion code? You enter it at checkout, where it is applied to your subscription — the prices above are the standard rates.