ByokVault - Your Keys, Every Model, One Price You Can See

chat

Bring your own provider keys or use ours. Every AI model from every provider in one catalog, with the exact price of every call shown before and after you spend.

Bring your keys. Or use ours. Your choice.

Connect your own provider keys or run on ours - and see the exact price of every AI call, from every model, in one place. No vendor lock-in. No hidden markup. No model-name guesswork. ByokVault is the encrypted key vault and unified model catalog underneath every AI feature on ToRun.

Connect a key · Read the BYOK guide


One vault. Every provider. No lock-in.

Most AI platforms quietly route you through their account, mark up the bill behind a flat "credits" number, and trap your data inside one vendor. ByokVault flips that. Connect your own OpenAI, Anthropic, Google, or other provider keys and ToRun routes your calls straight through them - you pay the provider directly, we charge a small, stated management fee. Don't have keys yet? Use ours and pay a transparent per-call rate. Either way you can pull the plug and walk away tomorrow, with full pricing in hand.

The same router that reaches a cloud provider can reach a model you host yourself - a local Ollama, vLLM, or LM Studio endpoint - so on-prem deployments run against their own models with no code change. Capability-first routing doesn't care where a model lives; it cares what the model can do.


A single catalog for every model

Behind the scenes, every model from every provider lives in one unified catalog - text, vision, image, video, audio, realtime, embeddings - each tagged with what it can actually do and what each capability costs. You never have to memorise model names or guess which one supports the thing you need. You ask for a capability ("vision", "long context", "realtime voice", "reasoning") and ByokVault picks the cheapest model and tier that delivers it. New models appear in the catalog without you changing a thing - the catalog can even sync live pricing so the numbers stay current on their own.

Capabilities go deep, not just broad: the catalog distinguishes things like single-image inpaint from multi-image composite, extended thinking from adaptive thinking, batch tiers from realtime - so routing filters on the exact shape of your request, not a coarse label.


Pricing you can audit, line by line

AI pricing is a moving target - providers price per million tokens, per image, per video minute, per audio second, per realtime minute, per search call, per embedding. ByokVault stores every one of those dimensions as its own time-stamped price, so the cost of a call is never a mystery. The moment you make a call, we snapshot the exact rate that applied - it can never be silently rewritten later. Every price change is logged in a change history. If you ever dispute a charge, the receipt is already written.

A single configurable margin is applied once, at billing time, by capability class - so the markup is one number you can point to, not a stack of hidden multipliers. The raw provider price is what we store; the marked price is what you see, and the gap is published.


Keys stay safe. Keys stay yours.

Your provider credentials live encrypted at rest in the vault. Every access is logged and alertable, and a credential resolution audit records which key answered each call - so you can always trace a charge back to the exact credential that served it. Keys are health-checked, so a revoked or rate-limited key surfaces before it breaks your workflow, and a provider outage is visible as a status rather than a mysterious failure. When you re-key, the old secret is gone.

BYOK works in tiers: route through your own active key when you have one, and fall back to a platform key when you don't - each path priced transparently, each call attributed to the fund source that paid for it.


What you get

  • Bring your own keys bi-key - connect your OpenAI, Anthropic, Google, and other provider keys - or a local model endpoint. Your calls route straight through them and you pay the provider directly; we only add a small, stated management fee.
  • One catalog, every model bi-collection - text, vision, image, video, audio, realtime, and embedding models from every provider in a single catalog, each tagged with its capabilities and the price of each one.
  • Capability-first routing bi-diagram-3 - ask for what you need, not a model name. ByokVault filters to models that support every capability your request requires - down to the exact input shape - then picks the cheapest for your tier.
  • Audit-proof pricing bi-receipt - every call snapshots the exact rate that applied - provider, capability, tier, and modality - so it can never be silently changed later. Every price change is logged.
  • Encrypted, health-checked vault bi-shield-check - credentials encrypted at rest, every access logged and alertable, with a credential-resolution audit and key health checks so a revoked key surfaces before it breaks your work.
  • Zero vendor lock-in bi-box-arrow-right - use your keys or ours, cloud or local, with full pricing visibility either way. Pull your key and self-host tomorrow.

Transparent margin 1.3-1.7x stated per capability · 12+ pricing dimensions tracked · 1 billing record per AI call - never more, never hidden.

Atomic billing: every AI call funnels through one BillingRecord with a full pricing snapshot and real-time usage transparency.
Atomic billing: every AI call funnels through one BillingRecord with a full pricing snapshot and real-time usage transparency.