pricing
Pay for proof, not promises.
One meter: depth. A run costs credits equal to how many rival agents you put on it, and you pay only when it meets the definition of good you set. A failed run costs nothing.
choose the roster depth · candidates / judges / integrators
- Billed only when the run meets its frozen Definition of Good. A failed run is free.
- Depth 2 and beyond can earn the cross-vendor Verified stamp when the run passes. Depth 1 is an honest, evidence-grade single reviewer.
- Same meter for Build and Review. No per-seat surprise on top.
If the result doesn’t land, the meter doesn’t move.
Every run freezes a Definition of Good before it starts. Meet it and the run bills its depth. Miss it and the run is free. That is the whole pricing model.
Free
kick the tires
forever · no card
- 30 credits a month
- Build and Review, both
- Depth up to 3/3/3
- The Receipt on every run
- $0.99 per extra credit
Pro
individual builders
per user / month, billed annually · $29 monthly
- 90 credits a month
- Depth up to 3/3/3
- Full cross-vendor jury
- Run history and receipts archive
- $0.66 per extra credit
Team
squads and orgs
per user / month, billed annually · $89 monthly · 2 seats min
- 300 pooled credits a month
- Depth up to 4/4/4
- Shared receipts across the team
- $0.55 per extra credit
Enterprise
high stakes
managed inference · private receipts
- Depth up to 5/5/5
- Managed or bring-your-own inference
- Local-first deployment
- SSO, audit, governance
- Per-verified-run billing available
Failed runs are free
Every run freezes a Definition of Good before it starts. If the result does not meet it, the run is not billed. You pay for verified work, not attempts.
Green is earned
Fusion Verified requires a passing objective gate and every hard criterion satisfied. It fails closed: when proof is missing, the stamp is withheld, never faked.
Early-access pricing
These are the launch numbers, and design partners help lock them. What will not change: the meter is depth, and unverified work stays free.
the fine print, answered
Pricing questions, answered plainly.
The basics
Is there a free plan, and do I need a credit card?
Yes. The Free plan is free forever and needs no credit card. You get 30 credits every month to build or review code however you like, at the same full-strength verification the paid plans use, up to a 3/3/3 run. It's a real working tier, not a stripped-down demo.
What's a "credit," in plain terms?
A credit is the unit you spend to run Fusion. Simplest way to think about it: a standard, fully verified run costs 3 credits. Lighter runs cost fewer; deeper ones cost more. Your plan includes a monthly bucket of credits, and you can buy more if you need them.
How much does one run cost?
You choose the depth, and the depth is the price. A quick single-reviewer pass is 1 credit; a full cross-checked verified run is 3. The price is the same whether you're building new code or reviewing existing code. On Pro, an extra credit is 66¢.
What does "charged only when it meets your definition of good" mean?
Before a run, you set what "done right" means: the tests that must pass and the criteria that must hold. If the run delivers that, you spend the credits. If it fails the gate, crashes, or errors out, you're not charged. You never pay for a run that didn't deliver.
How is this different from other AI review tools?
Two ways. Most charge per seat every month whether or not the tool caught anything; we charge per run, and only when the run actually delivers what you asked for. And they hand you one AI's opinion. A Fusion verified run uses multiple different-vendor models competing and cross-judging, which is why it can carry a Verified stamp.
Credits and depth
What does "depth" or "3/3/3" mean?
It's how many independent agents work your run: candidates / judges / integrators. A 3/3/3 run has 3 different models attempt the work, 3 independent judges score them blind, and 3 integrators produce the final. Then they cross-check each other. More depth means more independent verification, and it sets the credit cost.
Why pay for 3/3/3 instead of just 1/1/1?
1/1/1 is a single model reviewing the work. Fast and cheap, fine for a sanity check, but still one opinion. At 2/2/2 and up you get multiple different-vendor models competing and cross-judging each other blind, which is what can earn the Verified stamp: an independent, cross-checked result rather than one model's guess.
What's the difference between "Verified" and "evidence-grade"?
Verified means the run cleared its objective gate, the hard criteria in the Receipt are satisfied or explicitly waived, and independent judging has no unresolved split. A 2/2/2 run or deeper gives you the cross-vendor roster that can earn that stamp. A 1/1/1 run gives honest evidence from a single reviewer, but we won't call it Verified. We keep that word strict on purpose.
Do unused credits roll over?
On Pro and Team, yes, unused credits roll over up to 2× your monthly amount. Free-plan credits reset each month.
Why credits instead of a flat price per review?
Because runs come in different depths, you shouldn't pay the same for a 1-credit check as for a 3-credit verified run. Credits let you dial depth to the job and pay only for what you use, and they let a whole team share one flexible budget instead of buying rigid per-review packs.
Keys and models
Do I bring my own AI keys, and what does that cost?
On Free, Pro, and Team, yes. You connect your own AI accounts (Anthropic, OpenAI, Google, or an aggregator like OpenRouter), and those providers bill you directly. Fusion never marks up or resells tokens; you pay us for the verification, not the compute. On Enterprise, we supply the models so you bring no keys.
Which models should I use?
Our recommendation: if you want the best results, use the top frontier models. Fusion isn't about saving money on cheap models. It's about return on investment and trust in your codebase, where a competition among the strongest models produces a result you can stand behind. If the goal is just spending less, a free model elsewhere will do; if the goal is the best verified outcome, run the best models through Fusion.
Plans and billing
How does per-user pricing work, and why does Team need 2+ users?
Pro and Team are priced per person: $19/user/month on Pro (billed annually), $79/user/month on Team. Each user brings a bucket of credits (90 on Pro, 300 on Team), and on Team those credits pool across everyone. Team requires at least 2 users because it's built for collaboration; pooled credits, shared roles, and team controls only make sense with a team. Solo users are perfectly served by Free or Pro.
Monthly vs annual billing?
Annual is cheaper. Pro is $19/user/month paid annually vs $29 month-to-month; Team is $79 vs $89. Same product either way, annual just saves money for committing up front.
What happens when I run out of credits?
Keep going by buying extra credits at your plan's rate, 99¢ each on Free, 66¢ on Pro, 55¢ on Team, or upgrade. On Team, admins set how many credits each user can spend, so the shared pool stays under control.
Can I cancel anytime, and do you offer refunds?
You can cancel anytime. We don't offer refunds. Instead, when you cancel you keep access through the end of the period you've already paid for, then it stops. Enterprise cancellation follows the terms of your agreement.
How do I estimate what my team will spend?
Start from how many verified runs you expect. A standard verified run is 3 credits, and Team includes 300 credits per user per month. A 5-person team has 1,500 pooled credits, roughly 500 verified runs a month, before any extras. Heavy months add 55¢ per extra credit, and admins can cap per-user spend.
Enterprise and security
What is Enterprise pricing based on, and what's "managed inference"?
Enterprise is a custom contract sized to your usage: verified runs per month, the depth you need (up to 5/5/5), and requirements like on-prem deployment, governance, and SLAs. Managed inference means Fusion supplies the models, so your team brings and manages no API keys. One vendor, one invoice, billed per verified run.
Does Fusion see or store our code?
On the local-first plans, runs execute on your machine. The AI agents inspect the repository through the tools and accounts you already use, under the vendor agreements you already have. Fusion does not add a separate code-upload path to our servers. On Enterprise, code is handled inside your controlled environment under contract, with SSO, audit logs, and on-prem options. Details: the fine print on our about page.
Open source
Is Fusion free for open-source projects?
Yes, genuine open-source projects use Fusion free through our Open Source program. To qualify, the repo must be public under a recognized open-source license (MIT, Apache, GPL and similar), non-commercial, and you need to be the owner or an admin. Fusion's own platform stays closed-source; free for open-source projects is about who can use it, not about opening our code.
How do I apply for the Open Source program?
Send your name, the repository link, and its license to hello@fusioncoding.ai. We run a light eligibility check. Once approved, Fusion turns on free for that project. Your public pull requests start getting Verified Receipts at no cost. No sales call, no credit card.
The first receipt is on us.
Start free with 30 credits a month. Bring a repository and a task; leave with a verified pull request and the proof that it earned it.
get early access