Pricing

Pay for orchestration.
Not for compute.

Concurrent runs, schedules, multi-agent orchestration, and projects scale by plan — not the DevBoxes you launch, the terminal you live in, or the AI you choose. Bring your own machines and the compute is yours — or, if you'd rather not host, spin up Cloud workers and pay only for the compute you use.

Free

$0/mo

Bring your own machines and your own Claude subscription.

Starter

$9.99/mo

For solo developers getting started with AI coding agents.

Most Popular

Pro

$29.99/mo

For teams automating their sprint workflow with AI.
+$20/additional seat · up to 50 seats

Enterprise

$99/first user

For engineering orgs that need control, compliance, and scale.

Not sure yet? Book 15 minutes with the founders — no signup needed.

How We Think About Pricing

01
Tier on scale and automation, not core functionality. Starter is genuinely useful — not a crippled taste.
02
The terminal is always free. Across every plan, every DevBox, every machine. Non-negotiable.
03
The AI provider is your choice. We don't gate, upsell, or take a margin on model inference.
04
Don't charge per DevBox run. Usage-based pricing scares developers. We charge for concurrency limits instead.
05
BYOM keeps your costs low — or use ours. Your self-hosted workers are your compute, so we're the orchestration layer, not the GPU bill. Don't want to host? Cloud workers give you managed compute on demand, metered by the second.

Cloud worker compute

Need cloud workers too?

Pay per minute for Linux and Windows DevBoxes, from $0.06/hr. Billed from your wallet, available on every plan.

View Cloud worker pricing →

Frequently Asked Questions

What's a worker?

A worker is a machine that runs your DevBoxes. Two types: • Self-hosted (BYOM) workers — machines you own (your laptop, a spare desktop, a cloud VM). Free to connect; you pay only the subscription. • Cloud workers — VMs managed by ClusterCode, metered by the second from your wallet. Available on every paid plan; you pay for what you use.

WorkersCloud workers

What's a DevBox?

A DevBox is a hardware-isolated dev environment with a live terminal and your tools pre-installed. An agent works inside it — and you can connect to the same box, drive the terminal, or open a desktop. DevBoxes run on your workers, self-hosted or Cloud.

DevBoxes

What's the difference between launching a DevBox and starting a Run?

A launch gives you the machine; a Run gives you the result. Launching from a ticket prepares an isolated DevBox you own — repo cloned, branch checked out, ticket context written, agent CLI installed — and then stops. Nothing runs until you start the agent yourself in the terminal. A Run hands the ticket to an agent that works it to completion inside that same kind of DevBox, and opens a draft PR when it is done.

Launch vs Run

What are Runs?

A Run is on-demand agent work: paste a ticket or describe a task, and an agent spins up a DevBox, writes the code, runs the tests, and opens a PR — while you watch each step. Runs are driven by Nova and draw on your AI credits (or your connected AI subscription).

RunsRun on demand

What are Schedules?

Schedules are Runs on a cron — recurring automated work like a nightly sprint sweep, a scheduled dependency bump, or a morning PR-review pass. Free includes 1, Starter 3, Pro 15, Enterprise 60. The cadence floor scales by plan: Free runs at most 6 hours, Starter 1 hour, Pro 12 minutes, Enterprise 1 minute. Like Runs, they are driven by Nova.

SchedulesRecurring work

What are Loops? (Beta)

A Loop is a standing goal that runs until it's verified done. You describe the goal and the done-when checks; each iteration a maker agent does the work, deterministic gates run, and an independent verifier (a different model) judges the result — repeating until it passes. A Loop never auto-merges; it opens a PR for you. Free includes 1 active Loop, Starter 3, Pro 15, Enterprise 60. Loops is in beta.

LoopsCreate a Loop

What's Prototyping? (Beta)

A Prototype is one prompt tried by several engines at once. You write the brief and pick the engines (Claude, Codex, Copilot); each runs the same goal on its own branch, in its own container — so you compare real attempts side by side and promote the one you like. Promoting never merges anything — it opens a pull request from the winner's branch for you to review. Free includes 1 active Prototype with up to 2 engines each, Starter 2 · 3, Pro 10 · 4, Enterprise 40 · 6. Prototyping is in beta.

PrototypingCreate a Prototype

What are Nova and Nucleus?

The two agents built into ClusterCode: • Nova — the autopilot that drives Runs, Schedules, and multi-agent work, where several agents collaborate on one task (planning, coding, reviewing, testing). It consumes AI credits, or your connected AI subscription when wired up. Concurrent agents per run scale by plan: 3 on Starter, 10 on Pro, 25 on Enterprise. • Nucleus — the in-app assistant. Chat with it to launch DevBoxes, create schedules, inspect workers, and manage your workspace without leaving the page.

NovaNucleusMulti-agent runs

What's the Observatory?

The Observatory is a live map of your whole fleet — workers as suns, DevBoxes orbiting as planets, Runs and Schedules as moons — so you can see everything that's running at a glance and drill into any of it.

ObservatoryExplore the Observatory

What are Projects?

A Project bundles a repo with its container image, environment variables, and settings, so every launch, Run, and Schedule inherits the right setup automatically — no re-configuring each time. Free includes 1, Starter 3, Pro 25, Enterprise 60.

Projects

What's the wallet?

Every paid plan includes monthly AI and compute allowances. AI: $2 on Starter, $6 on Pro, $10/user on Enterprise. Compute: $2 on Starter, $6 on Pro, $10/user on Enterprise. Top up any time for extra usage; included allowances don't roll over.

AI credits

Do I need my own Claude API key?

Not required — the managed AI catalog (Claude, Copilot, Codex) runs through your wallet credit. Prefer to bring your own key or subscription? That's fully supported and reduces wallet usage for AI to near zero.

Runtime catalogBring your own subscription

What is BYOI?

BYOI — Bring Your Own Image — is a Pro+ feature that lets you use any custom container image from your own registry, beyond the managed catalog. Useful for internal dependencies, a fine-tuned model, or full control over the environment.

Build an image from a DevBoxCustom Containerfile

Can I self-host the entire thing — orchestrator included?

Self-hosted workers always run on your own machines on every plan — your code never leaves them. A fully self-hosted orchestrator (our control plane on your own infrastructure) is an Enterprise option that's coming soon; today the orchestrator runs on our cloud while your DevBoxes stay on your workers.

Workers

What happens after my 7-day trial?

When your 7-day Pro trial ends, you're not locked out — your workspace drops to the Free plan and keeps running on your own machine with your own Claude key. Resources above Free's limits (extra workers, schedules, longer run history) are trimmed back to the Free caps, but your account isn't deleted. No card is ever required for the trial, so there's no auto-charge — upgrade to a paid plan whenever you want the higher limits back.

How does Enterprise pricing work?

Enterprise is per-user with degressive pricing: $99 for the first user, $75 for users 2–10, $50 for users 11–50, custom pricing beyond 50. Each user includes a $10/mo AI allowance plus compute allowance.

Can I switch plans?

Anytime. Upgrades are immediate. Downgrades take effect at the next billing cycle. No contracts, no minimum terms on Starter or Pro.

Start building today.

7 days of full Pro access. No card, no commitment. See what your setup looks like before you pay for it.

Start your free trial →