Implemented in the project
- Unified model gateway with routing, API keys, policy, budgets, and audit
- Admin console and management API
- Local, Docker, database, and Cloudflare deployment paths
CinaToken / open-source, self-hosted AI gateway.
Route model access through one governed gateway.
CinaToken combines provider routing, API keys, model policy, budgets, accounting, and audit with local, Docker, database, and Cloudflare deployment paths. This page does not claim a managed-service SLA.
Product evidence / current scope.
What is verifiable today.
Status and capabilities below are tied to public project repositories. They do not imply a customer deployment, managed SLA, or production availability beyond the linked evidence.
Status: Open-source / self-hosted
The gateway repository includes local, Docker, and Cloudflare deployment paths. No managed-service SLA is claimed here.
Verified: · commit 3926095
Documented interfaces and integrations
- OpenAI-compatible Chat Completions and Responses APIs
- Image, audio, embeddings, Realtime, Claude, and Gemini routes documented by the project
- SQLite / D1, PostgreSQL, MySQL, Docker, and Cloudflare Workers
CinaToken
Example data generated by this site—not a customer screenshot or proof of a live deployment.
Input Responses API request
Control Route policy + budget check
Output Provider response + audit event
Capabilities.
A focused layer for the work your team needs to finish.
These are reference workflow patterns, not a promise that every capability is enabled in every release or deployment. Start with the smallest useful workflow and confirm current availability against the evidence above.
Expose unified model interfaces.
Connect applications through documented Chat Completions, Responses, image, audio, embedding, Realtime, Claude, and Gemini routes.
Issue controlled downstream access.
Keep provider credentials behind the gateway and manage user-facing API keys from the administration layer.
Select approved providers and models.
Route by explicit provider, model, task, availability, and policy configuration.
Apply recurring usage boundaries.
Use the project’s budget, accounting, status, and model-access controls for governed consumption.
Retain operational evidence.
Inspect routing, policy, usage, and accounting through the documented console and management APIs.
Choose a supported hosting path.
Run locally, in Docker, with supported databases, or through the documented Cloudflare deployment path.
Workflow.
Move one application through a governed model route.
A reference rollout starts with one protocol, provider route, downstream key, budget rule, and observable request.
01 / Choose the interface Select the documented protocol and confirm the application request shape.
02 / Configure a provider route Add the provider credential securely and restrict the route to the intended models.
03 / Issue a downstream key Create the user or application identity with its applicable model and budget boundaries.
04 / Verify usage and audit Send a test request, then confirm routing, response, accounting, and trace evidence.
Illustrative gateway policy.
Keep the implementation understandable.
Use explicit inputs, review points, and outputs so the system is easier to operate and change.
gateway "team" {
protocol = "responses-api"
route = "approved-models"
budget = "enforced"
audit = true
}Operating principles.
Built for practical adoption.
Self-hosted control.
Operators choose the hosting path, providers, keys, and data stores.
Protocol clarity.
Applications should use only interfaces documented by the active release.
Observable spend.
Routing, usage, budgets, and accounting should remain visible to operators.
Next step.
Map this product to one real workflow.
Tell us the work, systems, constraints, and review points your team already has. We will help identify a sensible starting scope.