Skip to main content
Tokios gives your local models a stable API endpoint at https://api.tokios.com. Use this page to choose the workflow you want, then jump into the setup guide that matches your client, backend, or deployment pattern.

Coding agents

Run Claude Code, OpenAI Codex, Aider, Cline, Continue, and other compatible coding tools against your own local models.

Local AI workstations

Expose models running on NVIDIA GB10, DGX Spark, AMD Strix Halo, Ryzen AI Max, Apple silicon Macs, and other workstations through one API endpoint.

Choose a local model

Pick a model and backend based on your workflow, hardware, and latency needs.

Benchmark a deployment

Compare local deployments with the same prompt, client, and hardware.

Model routing

Register multiple deployments and switch models by changing the model field.

​
Common workflows

​
How Tokios fits

Tokios separates three names that are easy to mix up:
  • The connector is the local tokios-connector process.
  • The upstream model id is the model name your local backend serves, such as an Ollama or LM Studio model id.
  • The deployment is the public model name your client sends in the model field.
Keep those names clear and most setup issues become easier to diagnose.
If you are starting from scratch, use the Quickstart. If you already know your client or backend, jump directly to the matching page above.