Start with secure AI agents in the cloud, priced per user. Move on-premises onto an annual capacity subscription sized to the work VDF AI actually performs — bought directly from VDF AI or through an authorized partner.
One annual subscription. Unlimited people. A governed capacity pool you buy up front — not a meter that bills you after the fact.
License VDF AI on your own infrastructure, or run it as a managed cloud subscription. Switch between the two to compare packages.
Capacity Licensing. Choose the annual capacity that matches the work VDF AI performs. Capacity is pooled across the licensed organisation, with transparent usage reporting and simple expansion when demand grows — and unlimited human users throughout.
Pilot-scale, single team
A first governed production workload with room to build alongside it.
1 production + 1 non-production
Departmental rollout
A department running VDF AI daily, with headroom for real adoption curves.
1 production + 1 non-production
Multi-team production
Several teams in production with separate development and staging estates.
1 production + 2 non-production
Multi-estate
Parallel production estates, higher throughput, and a full delivery pipeline.
2 production + 3 non-production
Organisation-wide
VDF AI as shared infrastructure across business units and regions.
4 production + unlimited non-production
Each step adds governed annual capacity on both meters, throughput headroom, and included environments. Deployments beyond the Enterprise bundle are supported under a negotiated enterprise agreement. Annual list pricing is provided on request or through your authorized VDF AI partner.
One subscription covering the platform, the people using it, and everything they build with it.
Capacity is measured on two meters, both published rather than buried in a schedule. The platform's own work — the orchestration, governance, and retrieval that make agents usable — never counts against either.
One unit of model input or output processed through VDF AI, counted once — including embedding tokens generated for retrieval indexing.
Model-intensive work: conversational assistants, retrieval-augmented generation, summarisation, code assistance, and agent reasoning.
One inbound request to VDF AI — an API call, a scheduled job execution, or a triggered workflow — counted once regardless of how much processing happens inside it.
Request-intensive work: integration logic, straight-through processing, event-driven automation, and high-frequency scoring.
Both meters run in parallel. A bundle reaches its limit when either allowance is reached, and capacity can be extended on either meter at the same cost — so there is nothing to predict in advance.
| Counted once | Included at no additional usage |
|---|---|
| Successful model input and output tokens | Included Orchestration, routing, and model selection |
| Embedding tokens generated for retrieval indexing | Included Governance, policy evaluation, and audit logging |
| Tokens processed by VDF Chat and VDF Agents | Included Retrieval from an existing index |
| One transaction per API call, scheduled job execution, or triggered workflow | Included Interactive sessions by your people — users are unlimited |
| One transaction per 100 records in a batch submission | Included Administration, configuration, and monitoring |
| Retries and regenerations you ask for | Included Failed requests, and automatic retries within 60 seconds |
Confirmed cache hits and VDF-generated technical retries are not counted either. Usage on both meters is drawn from actual provider-reported tokens and gateway-observed requests, and is visible in the same dashboard and monthly statement VDF AI reads — so each meter can be reconciled independently against your own logs. Image, audio, and video workloads require a written capacity schedule before production use.
Consumption within the permitted surplus above your allowance is not chargeable: no throttling, no invoice, no escalation. It is deliberate headroom for the weeks a workload runs hot, not a trap.
A usage dashboard, a monthly consumption statement, and staged alerts tell you where you stand long before an allowance runs low — and again on reaching the permitted surplus. No surprise invoices at the end of a quarter.
Beyond the permitted surplus, capacity is extended on the meter concerned and invoiced as incurred, and service continues. Only at the maximum surplus threshold recorded in your order form may throughput be reduced or a deployment suspended — and only after at least ten business days of written notice, and not at all if additional capacity is ordered within that period.
A portion of unused annual capacity carries into the following term when the subscription renews, so a conservative first year is not money written off.
Expansion Pack spend is credited in full against a mid-term upgrade to the next bundle. Growing faster than forecast costs you a conversation, not a penalty.
Before you commit
A defined capacity allowance, a throughput ceiling, and one non-production environment — enough to measure real consumption instead of guessing at it. The full pilot fee is credited toward an annual Capacity Bundle ordered shortly after the pilot completes.
A Capacity Bundle is required when a production deployment is reached through an API, service account, scheduler, integration, or another application. Human access is unlimited within the bundle — people are never the metered part.
A small interactive-only user licence is available for deployments where production API, service-account, scheduler, and integration access are disabled. If programmatic access is later enabled, there is a conversion window with full credit for the unused interactive term.
Capacity Licensing is how you buy VDF AI. These are the environments you can run it in — the licensing model is the same across all four.
Deploy inside your own AWS, Azure, or GCP tenant for cloud convenience with stronger data control and residency.
Run VDF AI entirely within your own data center for full sovereignty, auditability, and internal control.
Meet national data-residency and regulatory requirements with a governed, in-region deployment.
Operate in fully isolated environments with no external dependencies — defense and critical infrastructure.
Extend the subscription without changing the licensing model.
Raise the throughput ceiling for latency-sensitive and bursty workloads.
Disaster recovery, a second data centre, or a separately managed production estate.
Extra development, test, staging, or training environments.
Fully isolated operation with no external dependencies. Professional services required.
24×7 coverage, one-hour P1 response, a named technical account manager, and a defined escalation path.
Implementation, integration, migration, and model tuning, scoped against your project.
How the subscription behaves once it is signed.
Twelve, twenty-four, or thirty-six month terms. Multi-year commitments carry a subscription discount and lock your price for the committed term.
Organisations starting with a smaller deployment can agree a pre-defined annual step-up rather than over-buying capacity in year one.
Connected deployments produce signed monthly usage statements automatically. Air-gapped deployments produce quarterly offline statements. No separate usage audit is required where statements have been provided in good faith.
Capacity can be extended at any point in the term on whichever meter you reach first, priced the same either way — so you never have to predict which one your workload will exhaust. Expansion beyond the permitted surplus is bounded by the maximum surplus threshold recorded in your order form.
Estimate the annual capacity, peak throughput, and environments your workloads need.
Confirm the forecast through the pilot or against the logs of workloads you already run.
Select the bundle and agree top-up, support, and deployment options on the order form.
Sizing starts with a workload session: annual capacity, peak throughput, environments, and deployment mode. VDF AI software may be procured directly or through an authorized VDF AI partner, and implementation and professional services may be quoted separately by the delivering partner.
Looking for managed cloud plans?
Managed subscription plans we host and operate for you. Begin self-serve, expand into governed networks, and move to a sales-assisted Enterprise Cloud rollout when you need enterprise governance, SSO, and monitoring.
For teams starting with secure AI agents and connected knowledge workflows.
For teams building governed AI workflows, code assistance, and repeatable AI networks.
Everything in Starter, plus
Fully managed VDF AI with enterprise governance, SSO, and monitoring — without running the infrastructure yourself.
Everything in Professional, plus
Capabilities across the three Cloud plans. The Enterprise column also reflects what a self-hosted deployment includes, though on-premises is licensed on capacity rather than per user — see the .
| Capability | Starter | Professional | Enterprise |
|---|---|---|---|
| Licensing model | Per user | Per user | Per user (Cloud) / annual capacity pool (on-prem) |
| Secure AI workspace | Included | Included | Included |
| VDF AI Agents | Included | Included | Included |
| Adaptive, Manual, and Agent modes | Included | Included | Included |
| 20+ integrations | Included | Included | Included |
| Custom agents | Included | Included | Included |
| Role-based access control | Included | Included | Included |
| VDF AI Networks | Basic workflows | Networks Core | Unlimited Networks |
| Visual network builder | Limited | Included | Included |
| Network limit | Starter workflows | Up to 50 networks | Unlimited |
| Private RAG | Standard | Advanced | Enterprise / on-prem |
| MCP server support | — | Limited / add-on | Included |
| Model routing | Basic | Advanced | Policy-driven |
| Governance controls | Standard | Advanced | Enterprise-grade |
| Audit logs | Basic | Advanced | Immutable enterprise logs |
| Monitoring and cost analytics | — | Included | Advanced |
| Deployment | Managed cloud | Managed or private cloud | Cloud, on-prem, sovereign, air-gapped |
| Custom integrations | Add-on | Add-on | Included / custom scope |
| Single sign-on (SSO) | — | — | Included |
| Support | Standard | Priority | Enterprise SLA |
| Training and onboarding | Self-serve | Guided onboarding | Dedicated rollout program |
Prefer to run VDF AI inside your own environment?
VDF AI Cloud plans are available directly from VDF AI. Enterprise deployments can be purchased direct, subscribed through AWS Marketplace into your own AWS environment, or procured through an authorized VDF AI partner — whichever route fits your location, implementation needs, and procurement model.
Start VDF AI Cloud self-serve, or work with our team for a sales-assisted rollout and enterprise proposal.
Deploy VDF AI Enterprise from AWS Marketplace into your AWS environment, subscribed through the account your organization already runs on.
Purchase VDF AI through an authorized VAR or system integrator who delivers licensing alongside implementation and local support.
Partner program benefits and eligibility for VARs and system integrators are described on our Partners page →
Most AI platforms bill you after the fact, in units you cannot forecast: credits, executions, agent runs, seats, add-ons, and hidden limits. VDF AI inverts that. You size and buy capacity in advance at a fixed annual price, you know exactly what does and does not consume it, and you keep deciding which models run, where, and how data is handled.
On-premises deployments run on a governed capacity pool you commit to up front at a fixed annual price. You are not invoiced after the fact for whatever your agents happened to do last month.
Administrators, operators, business users, and developers are unlimited. Adoption is never rationed by seat count or held back by a licence review.
What counts toward capacity — and what never does — is published, not buried in a schedule. Usage is drawn from provider-reported tokens in the same dashboard VDF AI reads.
Use OpenAI, Anthropic, Gemini, Mistral, Llama, or your own on-prem models. VDF AI routes on quality, cost, latency, and policy.
Enterprise licensing supports managed cloud, private cloud, on-premises, sovereign cloud, and air-gapped environments.
Distinct from the licence add-ons on the On-Premises tab: these are delivery engagements, scoped and quoted separately from any subscription — by VDF AI or by the authorized partner running your project.
Workspace setup, governance configuration, team training, and first workflow deployment.
Connect VDF AI to internal tools, private APIs, databases, document systems, and proprietary workflows.
Configure audit trails, approval gates, model policies, access controls, and compliance reporting.
Set up secure retrieval across internal documents, tickets, repositories, and structured data.
Analyze your AI stack and find where routing, on-prem deployment, or consolidation can reduce cost.
They are priced on different mechanics. VDF AI Cloud is a managed subscription we host and operate for you, priced per user (Starter, Professional, and Enterprise Cloud). On-premises Capacity Licensing is an annual subscription you run in your own environment — private cloud, on-premises, sovereign cloud, or air-gapped — priced on the governed capacity your workloads need rather than on headcount. Human users are unlimited under a Capacity Bundle.
Yes. VDF AI software can be purchased directly from VDF AI, through an authorized VDF AI partner, or — for the on-premises enterprise edition — via AWS Marketplace. Partners are VARs and system integrators who deliver licensing alongside implementation, integrations, training, and local support, depending on your region and preferred procurement model.
Yes. VDF AI Enterprise, the self-hosted edition of the platform, is listed on AWS Marketplace as a container-based product. You subscribe through your existing AWS account, deploy into your own AWS environment, and keep billing consolidated on your AWS invoice. Cloud plans and non-AWS deployments continue to be purchased directly from VDF AI or through an authorized partner.
No. The VDF AI software licence is separate from implementation and professional services. Services are scoped separately and may be delivered by VDF AI or by the authorized partner delivering your project.
Not as consumption billing. VDF AI Cloud is priced per user. On-premises deployments are licensed as an annual capacity subscription: you pre-purchase a governed capacity pool for a fixed annual price, with unlimited users. It is a commitment you size in advance, not a bill that arrives after the fact, and capacity is measured on two meters — tokens and platform transactions — with unlimited people either way. Model-provider charges and the infrastructure used to run models remain separate and depend on the deployment and models you choose.
Two meters. On the token meter: successful model input and output tokens, embedding tokens generated for retrieval indexing, tokens processed by VDF Chat and VDF Agents, and any retries or regenerations you request. On the transaction meter: one count per inbound API call, scheduled job execution, or triggered workflow, and one count per 100 records in a batch submission. Everything the platform itself does is included at no additional usage: orchestration, routing and model selection, governance, policy evaluation and audit logging, retrieval from an existing index, and administration, configuration and monitoring. Interactive sessions by your people, confirmed cache hits, failed requests, and automatic retries within 60 seconds of a failure are not counted at all.
Ordinary variation is free: consumption within the permitted surplus above your allowance is not chargeable, with no throttling and no escalation. Beyond that, capacity is extended on the meter concerned and invoiced as incurred, and your operations continue uninterrupted. Only at the maximum surplus threshold recorded in your order form may VDF AI reduce the throughput ceiling or suspend the affected deployment — and then only after at least ten business days of written notice, and not at all if additional capacity is ordered within that period. Separately, requests above your sustained throughput ceiling are rate-limited so that service continues; they are never rejected because of the ceiling alone.
One inbound request to VDF AI — an API call, a scheduled job execution, or a triggered workflow — counted once regardless of how many rules, integrations, steps, or model calls run inside it. A batch submission counts one transaction for every 100 records or part thereof, a scheduled job counts per execution rather than per schedule defined, and a streamed response counts once at the request that initiated it. Failed requests, and automatic retries within 60 seconds of a failure, are not counted. The transaction meter exists because request-heavy deployments — integration hubs, straight-through processing, high-frequency scoring — consume very few tokens per request, and a token-only meter would misprice them badly.
No. Administrators, operators, business users, and developers are unlimited within a Capacity Bundle. A small interactive-only user licence exists for deployments where production API, service-account, scheduler, and integration access are disabled; if programmatic access is later enabled, there is a conversion window with full credit for the unused interactive term.
Yes. A 90-day pilot includes a defined capacity allowance, a throughput ceiling, and one non-production environment. The full pilot fee is credited toward an annual Capacity Bundle ordered shortly after the pilot completes, so a pilot that proves the case is not a sunk cost.
Air-gapped deployments produce quarterly offline usage statements; connected deployments produce signed monthly statements automatically. Where signed statements have been provided in good faith, no separate usage audit is required.
It means users can create and use agents without per-agent fees. Fair-use safeguards apply to protect platform reliability and prevent abusive or excessive workloads.
Yes. VDF AI is model-agnostic. You can use commercial models, open-source models, or private models running in your own infrastructure.
Yes. VDF AI is designed for organizations that need data sovereignty, access controls, audit logs, private RAG, model governance, and deployment control.
By Capacity Bundle. Each bundle sets an annual allowance on both meters — tokens and platform transactions — along with the sustained throughput ceiling, the number of integrated systems, and the production and non-production environments included. The bundle binds on whichever allowance is reached first, so sizing starts by estimating both. Price is a function of the bundle you select plus term length, support level, deployment mode, and any add-ons. Annual list pricing is provided on request or through your authorized VDF AI partner, and implementation and professional services are quoted separately by the delivering partner.
Microsoft Copilot is optimized for the Microsoft ecosystem. VDF AI is a model-agnostic orchestration and governance layer that can run across your tools, models, knowledge sources, and infrastructure.
Start with secure AI agents in the cloud today. Move to an on-premises capacity subscription when control and compliance become non-negotiable — unlimited users, a pool you size in advance, bought directly or through an authorized partner.
Tell us what you’re trying to achieve—governed AI Networks, enterprise RAG, deep integrations, or on‑premise deployment. We’ll help you map the right architecture, security posture, and rollout path. If you’re moving beyond AI pilots and need scalable, auditable execution, reach out—our team is ready to help.