Stable model names

Use stable aliases such as tekizai/frontier, tekizai/balanced, tekizai/agentic, and premium tekizai/fusion. Available models depend on your hosted account or the providers configured on your router.

Works with the Responses API

Hosted, local, and Tailnet setups support OpenAI-compatible chat completions and the Responses API, including Codex profiles with wire_api = "responses".

Try another eligible route

Hosted requests can retry on eligible TEKIZ.AI capacity. A local or Tailnet router can retry among the providers and models you configure. A response stays on one route once streaming begins.

Choose a model that fits the request

TEKIZ.AI looks at the requested profile, tool use, context, provider health, and account limits before choosing a model.

Match the task

Tool calls, code/debug language, long context, vision hints, and requested profiles can push traffic toward models that fit the task.

Keep keys private

When you run the router, provider keys and sign-in profiles stay on that router. A hosted TEKIZ.AI key accesses models available to your account.

Deploy where agents run

Use https://api.tekiz.ai/v1 for hosted access, http://127.0.0.1:8790/v1 for local agents, or http://<tailnet-host>:8790/v1 for private Tailnet routing.

Codex profile

# ~/.codex/config.toml - add this provider once
[model_providers.tekizai]
name = "TEKIZ.AI"
base_url = "https://api.tekiz.ai/v1/"
env_key = "OPENAI_API_KEY"
wire_api = "responses"

# ~/.codex/tekizai-frontier.config.toml
model_provider = "tekizai"
model = "tekizai/frontier"

Plain codex keeps using ChatGPT. Select this provider explicitly with codex --profile tekizai-frontier.
Local and Tailnet Codex profiles use the same model profile with port 8790 on the router host.

Codex and code agents

Point Codex, OpenClaw, Cursor-style tools, and OpenAI SDK clients at one endpoint while preserving tool-call and Responses API compatibility.

Switch from another gateway

For hosted access, point an OpenAI-compatible client at https://api.tekiz.ai/v1 and use a model available to your account. Run the local router when you need your own provider keys or local models.

Keep provider access under your control

Keep provider credentials on a local server, private Tailnet host, or your own cloud when you run the router there. Hosted API keys use the models available to your TEKIZ.AI account.

Check availability

You can check the public feature status at https://api.tekiz.ai/features/agent-native. Public hosted model traffic is enabled only when the edge reports verified capacity.

Choose a setup

Send a request to the hosted API, run the router on your own machine, or use the calculator to compare estimated costs and plan request limits.