VENDOR-NEUTRAL AI · SELF-HOSTED CONTROL
Bring AI into your product.Keep control at the center.
Give your applications access to local and cloud AI through one controlled middle tier. Choose your providers, protect access, and turn quality review into a foundation for better customer experiences.
Evaluation & purchase · October 12, 2026
Ollama Cloud
Your choice of models
Provider freedomChoose models for the work.
Security by architecturePut control between apps and AI.
Evidence-led improvementReview answers. Refine the experience.
THE MISSING MIDDLE TIER
Build your product. Bring the AI controls with you.
Connecting to a model is only the beginning. Your institution also needs to decide who can use it, what information it receives, how spending is bounded, and how its answers will be reviewed. AI Router brings those responsibilities together, so your team can focus on the experience customers came for.
01 / FREEDOM TO CHOOSE
Many models. One place to manage them.
Bring local Ollama models and cloud providers such as OpenAI, Anthropic and Ollama Cloud into a shared administration experience. Organize models into service tiers and control which applications can use them—without building your product around a single AI vendor.
- Keep provider credentials on the server, out of customer browsers.
- Match access to each application’s authorized models and tiers.
- Evaluate models before approving them for specific product tasks.
02 / CONTROL THAT GOES DEEPER
Sensitive work deserves more than an API key.
Security is built into the boundaries between people, applications, administrators and providers. Each gets the authority it needs, with checks where decisions are made.
Separate the responsibilities
Administration and inference use separate interfaces. An administrator login does not automatically unlock captured conversations.
Protect privileged access
Invite-only accounts, multi-factor authentication and fresh verification for sensitive actions help keep powerful controls in the right hands.
Keep credentials contained
Provider keys are encrypted and write-only in administration. Rotation and scheduled validity keep credential management centralized.
Recheck permission as work moves
Authorization is checked at admission and dispatch. Revocation stops output and continuation instead of relying only on the initial login.
Put bounds around consumption
Client and deployment budgets, bounded queues and token reservations keep AI requests inside configured resource limits.
Keep tools under application control
Your application authorizes every tool operation. Model output never becomes permission to access a customer’s financial information.
03 / A PATH TO BETTER ANSWERS
Make quality something you can examine.
A fluent answer is only a starting point. Review a consented, captured turn against its recorded evidence, inspect structured findings, and record a human judgment. Give your team a concrete basis for improving instructions, model choices and customer experiences.
SIX DIMENSIONS OF QUALITY
- Grounding
- Arithmetic
- Assumptions
- Tool use
- Domain behavior
- Clarity
Current manual cloud audits assess a selected turn, with separate permission and consent. Scores support human review; they do not certify accuracy. Broader conversation review and comparison are planned.
Permission to use AI is not permission to record everything.
Capture, content review and external audit have distinct controls. Optional captures are encrypted, access is logged, and retention and withdrawal are enforced. Reviewers verify their identity before opening protected content.
BUILT FOR WEALTHPROOF. CREATED TO BE USEFUL BEYOND IT.
Responsible AI should be within reach.
We’re building AI Router as a key part of WealthProof’s future: helping people explore their financial accounts with clear evidence, visible assumptions and thoughtful oversight. We’re preparing it for like-minded institutions that want to bring the same discipline to their own products.
Our financial clarity principlesIncorporating AI while focused on the rigor and standards of the fintech industry.
WealthProof remains in development. Financial-account integration and further evaluation capabilities are planned; this is our direction, not a claim of a released financial service.
SERIOUS CAPABILITY. ACCESSIBLE ENTRY.
A shared foundation. An institution-sized price.
Self-host AI Router on your infrastructure, with your provider accounts. Every planned organizational tier includes the same core routing, security and audit capabilities, with unlimited hosts for the covered legal entity. Businesses with annual revenue of $500,000 or less qualify for a free certificate.
- Unlimited hosts per covered entity
- Free eligible business and nonprofit tiers
- 15% off a prepaid three-year term
Software license only. Infrastructure and provider costs are separate. Public images, certificates and client SDKs are being prepared for launch. Eligibility and license scope are confirmed with Dezota before issuance.
A few things worth knowing
Does vendor-neutral mean every model is interchangeable?
AI Router gives applications a shared routing boundary across its supported adapters. Models still differ in capability and terms, and require evaluation for their intended work. Routing does not silently fall back to another provider. Current quality audits use an OpenAI cloud auditor.
Who runs the infrastructure?
Your team operates the deployment and provider accounts. Planned software support covers product bugs and image or upgrade defects. Managed hosting, customer DevOps and an uptime or support response-time guarantee are not included.
How do we get started?
Contact Dezota for a product walkthrough and a discussion of your use case and license tier. Evaluation and purchase open October 12, 2026, with a 30-day organizational evaluation, extensions by agreement and no automatic paid conversion. We’ll agree evaluation dates and permitted use with you before issuing a certificate. Planned client SDKs cover JavaScript/TypeScript, Go, Rust and Python under MIT; the server will remain proprietary.