Procurement won't approve managed AI. IT won't approve closed-source. Your data can't leave the EU. Run the full WinGrants AI engine on your own infrastructure — open-source models, your keys, your audit trail.
If you're reading this page, you've already tried to procure something. 3 things keep blocking the sign-off.
EU research institutions and national agencies have a duty of care over consortium IP. SaaS contracts that put your data in someone else's database fail the audit — before the conversation even reaches the grants office.
OpenAI, Anthropic, Google — closed weights, opaque pipelines, US jurisdiction. Their TOS says "we don't train on your data," but there's no audit hook to inspect, no logs you can pull, and no way to prove your sensitive proposal IP didn't leak into the next model release. Once your data has crossed the Atlantic, the CLOUD Act doesn't require them to tell you.
A €4M EIC proposal is multiple partners' work. Putting it through a multi-tenant database that other organisations also use is a category of risk most consortium leads can't get signed off — rightly so.
Most teams don't need on-prem. Here's the honest read on where each plan fits — so you don't over-buy.
One researcher, one proposal at a time. EU-hosted SaaS, no procurement gate, pay only when you commit to submitting.
3–30 researchers in a grants office, RTO, or consultancy. Fixed annual licence; we ingest your past wins and codify your methodology.
Institutions where IT can't sign off on tools they don't host. Air-gap-capable, on-prem or your sovereign cloud, open-source models pinned to versions you choose.
WinGrants AI runs on open-source LLMs only — Llama, Qwen, Mistral, DeepSeek, Gemma, gpt-oss. Each module routes to the model that's strongest for the task: long-context for proposal review, reasoning for evaluation, multilingual for translation. No commercial API calls, no surprise pricing, no model deprecation drama.
3 deployment topologies — managed EU cloud, dedicated private cloud, on-premise (incl. air-gapped). Pick the one that matches your IT department's risk posture. Identity, encryption, and audit logging integrate with the stack you already run.
Enterprise isn't a bigger licence — it's a different relationship. Your IT team gets a deployment they can audit, on infrastructure they control. Your grants office gets the same hands-on rollout as AI Enablement. And you get the founders behind both.
The full WinGrants engine deployed on your own infrastructure — on-premise, your EU cloud tenancy, or air-gapped. Unlimited seats across every PI, grant officer and project manager.
2–4 week founder-led ramp-up with your IT and grants office in the room. We co-deploy with your engineers, configure security policies live, and train your researchers on real cases before we leave.
The platform runs on open-source models you control — Llama, Qwen, Mistral, DeepSeek — pinned to versions your team chooses. BYOK or HYOK. No closed-model vendor in the chain.
Shared Slack channel with the founding team for both your IT and your grants office. Async, no tickets, no support tiers — including roadmap input. Quarterly tuning sessions and ongoing EU funding advisory included year-1.
Universities and RTOs kept telling us — we love the platform, but our IT can't sign off on a tool we don't host. Enterprise is the answer to that conversation.
Personalise the inputs to match your institution. €30k/year licence anchor + €5k setup (scoped on the call) — uncapped use on your own infrastructure.
Caveat: Rough estimate (±10%). The €30k licence figure is the floor — Enterprise pricing is scoped to org size, integration scope, and SSO/audit complexity. The discovery call gives you a firm number.
Your data never leaves your perimeter. Contractually written into the DPA, audited annually by your IT, and verifiable in our open-source codebase. No sub-processors, no telemetry, no exceptions.
Every model in the chain is open-source. We will never swap in a closed-model dependency without your written approval.
On termination you keep the Docker bundle, your database export, every custom prompt, and the model weights. Stand up an offline instance forever.
We iterate until your team is producing submission-ready drafts — no cap on rounds, no clock. If we can't get you there under our Fair-Use Policy, the unused portion of the licence is refunded.
4 weeks, founder-led, alongside your IT team — we can compress to 2 weeks if your data and security review are ready. Week 1 — provisioning on your infrastructure (Kubernetes, Docker Swarm, or bare metal) and security policies. Week 2 — integration with your IdP, CRIS/DMS systems, and audit logging. Week 3 — ingestion of your past winning proposals and methodology codification. Week 4 — team training with your PIs and grant officers on real cases.
The platform currently routes between Llama 4 Maverick, Qwen 3.6, Mistral Large 3, DeepSeek V4, Gemma 4 and gpt-oss-120B depending on task — long-context for review, reasoning for evaluation, multilingual for translation. You can override the routing entirely if you have institutional preferences (e.g. only Llama, only EU-developed Mistral). Models are version-pinned; upgrades require your written approval.
The deployment lives wherever your IT chooses — including non-EU regions. We have customers in Switzerland, the UK and Norway running entirely on national-cloud infrastructure. The DPA reflects whichever jurisdiction applies; we are ISO/IEC 27001:2022 certified regardless.
We notify you in writing, propose a replacement open-source model from our supported list, and your IT signs off before we deploy it. The old model stays available in your tenancy until you choose to retire it. No silent swaps. No deprecation drama mid-proposal.
Yes. We ship the platform as a self-contained Docker Compose bundle with weights for all open-source models included. Zero outbound network calls required. Updates delivered as offline bundles signed with our private key, verified against a public root of trust your IT controls.
Shared Slack channel with the founding team for both your IT and your grants office — no tickets, no support tiers, no AI bots. Quarterly tuning sessions in year 1 to keep prompts and scorers sharp. Custom SLA available for organisations with formal uptime requirements.
Fixed annual licence — not per-seat, not per-proposal. From €30k / year + €5k one-time setup, scoped to organisation size and deployment shape. No public price list — we'll quote on the discovery call.
Usually a CISO or founder on our side. Mutual NDA before any technical detail. 30 minutes, no sales pressure.
Bring your IT lead and your grants director. We'll walk through the deployment shape, residency requirements, and the open-source model chain. No sales reps, no follow-up emails — just the founding team.
Book the discovery call →