One platform. One price per contributing developer.
Everything on this site — code review and the gateway, plus the microVM fabric, TensorLine tracking, and the TensorCode agent harness when they ship — is one product at one price: €35 per contributing developer per month, billed monthly via Stripe, from your first developer. No minimum seats, no yearly prepayment, and volume discounts are published below. Model usage is paid to your providers directly, at their prices, with zero markup from us.
Platform
Everything includedThe complete suite, on your infrastructure — or hosted by us in the EU. No minimum seats — the price scales from your first developer.
- TensorCheck PR reviews on unlimited repositories with Check Runs
- TensorCode agent harness (coming soon — included when it ships)
- TensorGate LLM gateway with zero token markup and OpenBao key custody
- TensorLine tracking with 20 native MCP tools (coming soon — included when it ships)
- TensorOrigin sovereign Git storage engine on immutable S3
- TensorBoot microVM sandboxes (coming soon — included when it ships), with capacity scaling with your seats
- Enterprise governance: AI readiness scoring, skills registry, SIEM export
- Cryptographic hash-chained audit trail (tp/audit/v1) with external witness
- Deploys on your premises or VPC; disconnected operation within an offline grace window (full air-gap on the roadmap)
- Or vendor-hosted in the EU by us — EU data residency from day one
- A data processing agreement on every contract — self-hosted mode sends us licensing telemetry only
- SSO via OIDC, SCIM, SIEM export, audit retention per your policy (SAML on the roadmap)
A data processing agreement accompanies every contract — see all legal documents.
Volume bands
Monthly platform fee by contributing developers, billed via Stripe. Discounts are already in the numbers — no negotiation required to get them, and a band change takes effect at the next monthly cycle.
| Developers | Rate | Monthly example |
|---|---|---|
| 1–99 | €35 / dev / month | 50 devs → €1,750 / month |
| 100–249 | ≈ €33.33 / dev / month | 100 devs → €3,333 / month |
| 250–499 | ≈ €28.33 / dev / month | 250 devs → €7,083 / month |
| 500+ | €25 / dev / month | 500 devs → €12,500 / month |
| The €25 rate applies to every organisation from 500 developers up, with no upper bound. Public-sector terms or special contracts? Talk to us | ||
Sandbox capacity: included
MicroVM sandbox capacity for TensorCheck PR reviews is included with the platform and scales with your contributing developers each monthly cycle — there is nothing extra to buy and nothing to pre-order for those review sandboxes. As your team grows, capacity grows with it; a larger cycle is simply priced at the volume band the new count lands in. Background cloud agents dispatched via TensorCode are metered separately, per agent-hour.
Cloud agent compute: pay as you go
From €0.12 / hrAutonomous background agents dispatched via TensorCode run inside dedicated TensorBoot microVMs. Sandbox capacity for PR reviews is included with your platform subscription; extended cloud agent compute is billed per agent-hour metered to the second, and completely free when self-hosted on your own hardware.
Model usage: at cost, always
TensorGate calls providers with your keys, at your prices. We add zero markup and never resell inference — the platform fee is our only revenue, so cutting your model bill is our job, not our loss.
A 100-developer company, in full.
| Platform — 100 contributing developers | €3,333 / month (≈ €40,000 / yr) |
| Model usage — your providers, your keys, zero markup | your provider bill, at cost |
| Sandbox capacity, upgrades, support — included | €0 |
That's the entire price. No per-product fees, no per-repository fees, no token resale, no surprise overages: each monthly billing cycle the platform snapshots the trailing-90-day contributor count and bills exactly that count through Stripe, and nothing ever stops working in between. The count is computed from unique non-bot git authors and is visible to you live in your admin console — the same number the invoice uses. The full mechanics are in the billing documentation.
The platform is pre-launch. Early access is free, includes the whole platform, and requires no payment details — request early access and we'll onboard your team.
Public-sector terms, special contracts, or interested in the air-gap roadmap? Talk to us.
Common questions.
What exactly is a contributing developer?
A unique, non-bot git author whose work the platform touched in the trailing 90 days — someone whose pull requests TensorCheck reviewed or who ran agents through the suite. Read-only console users, managers, and service accounts are free. The metric is defined in the contract and visible live in your admin console, so the number you see is the number on the invoice.
What happens if our team grows or shrinks?
Nothing blocks or warns you mid-cycle — the next developer's pull request gets reviewed like everyone else's, and work keeps running. Each monthly billing cycle the platform takes an immutable snapshot of the trailing-90-day contributor count and bills that count through Stripe. If your count grows or shrinks past a volume band, the new band takes effect at the next monthly cycle, in both directions, and the whole organisation is priced at the band the count falls in. A cycle with zero contributing developers invoices €0 and keeps your subscription and access active.
What do we pay for model usage?
Your providers' prices, and nothing more. TensorGate routes calls through your own API keys directly to EU provider endpoints or to models on your own GPUs. We add zero markup to tokens and never resell inference — our revenue is the platform fee, so our incentive is to reduce your model bill, not grow it.
How does cloud agent compute pricing work?
Review sandboxes for TensorCheck are included with your platform subscription. When running autonomous background coding agents via TensorCode, execution takes place inside dedicated TensorBoot microVMs billed at €0.12 per agent-hour (Standard: 2 vCPU, 4 GB RAM) or €0.24 per agent-hour (Performance: 4 vCPU, 8 GB RAM), metered to the exact second. If you deploy TensorPlane on your own infrastructure, agent compute is completely free (€0.00) — you provide the hardware, and we never charge for compute hours.
So what does a 100-developer company actually pay in total?
€3,333 per month for the platform (100 developers in the 100–249 band at ≈ €33.33 each, ≈ €40,000 / yr), plus your own model bill paid directly to providers at their prices — TensorGate's routing and caching are designed to reduce that bill, and its metering shows you exactly what they save. Sandbox capacity for TensorCheck PR reviews is included and scales with your contributing developers; background cloud agents dispatched via TensorCode are billed per agent-hour, metered to the second, and free when self-hosted. There are no other fees beyond those.
Is there a trial, and how does it start?
Yes — every hosted organisation starts with a 14-day trial that needs no card. You get the full product during the trial, and a subscription is required to continue after it. The 14-day term is fixed from activation, and there is one trial per GitHub organisation. See the billing documentation for the mechanics.
Is early access really free?
Yes. While the platform is pre-launch, early access is free and no payment details are collected. Onboarding is hands-on — we set each deployment up together. The prices on this page apply at general availability.
Where does our code go?
In the default self-hosted deployment the suite runs on your premises or in your VPC, and your code and model traffic never transit us. Model calls travel exactly one hop: from your own TensorGate directly to the EU endpoints of the providers you choose, under your contracts and keys — no TensorPlane servers and no US intermediary in that path. What we do receive in this mode is licensing telemetry — sign-in (SSO), entitlement and configuration refresh, heartbeats, revocation acknowledgements, signed usage records, with usage records dimensioned by organisation, team, user, repository, project, model, provider, skill, task type, time, which the control plane needs for licensing and billing — covered by the data processing agreement that accompanies every contract. If even one hop is too far, self-hosted models behind TensorGate keep inference on your own hardware; permanently air-gapped operation is on the roadmap, and today the platform tolerates disconnection within an offline grace window. If you choose the vendor-hosted option instead, the trust boundary changes and we say so plainly: your code is then processed on infrastructure we operate in the EU, under the same DPA — model calls still go through your keys with zero markup.
Can we cancel at any time?
Billing is monthly, with no yearly term and no fixed-term commitment. Once you are on a paid subscription you cancel yourself from the Stripe billing portal, and the cancellation takes effect at the end of the current monthly cycle — there is no yearly term to wait out. During early access there is nothing to cancel: access is free and no payment details are collected.