Solutions

Token Gear for operators, enterprises and AI builders

What each kind of customer can use now, and what is planned. Invite-only capabilities are offered to customers under an offline contract while we finish testing with real providers; planned ones are not available yet.

AI SaaS & integrators

Build on one API

Ship your product on one OpenAI-compatible API, with each request’s cost traceable.

  • Invite-onlyYour SDK, one new addressChange the base URL; an Idempotency-Key on each request keeps a retry from being charged twice; the compatibility matrix says which tools work.
  • Invite-onlySafe retriesA retry with the same Idempotency-Key and body gets the first answer again and is not charged twice.
  • Invite-onlyPer-request chargesEach answer carries its request id, so its tokens and charge can be matched to your own records.
  • PlannedPlaygroundTry models and parameters in the console before writing code.
  • PlannedYour own provider keysBring your own provider keys, billed apart from the platform’s procurement.

Enterprise AI governance

Govern your teams’ AI use

Put your teams’ model calls behind one gateway, with access, spend and data handling controlled in one place.

  • Invite-onlyKeys, budgets and allow-listsPer-key budgets, rate limits and model allow-lists; keys are shown once and stored as hashes.
  • Invite-onlyData constraintsRequirements on regions, training and retention, which can only narrow where a request may go.
  • Invite-onlyOrganizationsShared balances and members for a team or a company.
  • PlannedProjects and applicationsProjects and applications inside an organization, each with its own keys and records.
  • PlannedLayered budgets and alertsBudgets and alerts per organization, project, application and key, the strictest applying.
  • PlannedEnterprise sign-inOIDC and SAML single sign-on.
  • PlannedPrivate deploymentToken Gear run in your own environment: contact us to discuss it.

Operators & cloud providers

Run a model service

Offer models from several providers to your own customers, with one sales price per model and procurement kept apart.

  • Invite-onlyOne API, one price per modelAn OpenAI-compatible API for your customers, at one sales price per model whichever provider serves a request.
  • Invite-onlyRouting among admitted deploymentsEach request goes to an eligible deployment by capability, data constraints and procurement cost, with failover before any output is sent.
  • Invite-onlyRecords for every requestEach request, its upstream attempts, usage and charge are recorded, for customer questions and supplier reconciliation.
  • PlannedHierarchy and white-labelGroup, regional and channel organizations, internal settlement and white-label branding.

From tenant requests to Token Engines

Token Gear sits between tenants and NCP domains: it authenticates each request, applies the account’s limits and data constraints, picks an eligible deployment and meters the result. Each NCP’s Token Engine runs the inference.

  1. Tenants and their usersEach tenant’s requests, from its own users
  2. Token GearAI gateway · identity, routing, limits, data constraints and metering
  • NCP AToken Engine · LLM hosting · GPU / compute
  • NCP BToken Engine · LLM hosting · GPU / compute
  • NCP CToken Engine · LLM hosting · GPU / compute
An illustration. Requests go down; answers and usage come back over the same API.

Talk to us about joining →

Talk to us

Sign up, then contact FreshToken to sign an offline contract; an administrator adds your credit.