Documentation snapshot · CLI 7.30.1 · SDK 2.6.53. Desktop uses the runtime bundled with its own release. Commands and examples are preserved from the English source.

Connect your first model provider

Status: Active Scope: current-state Last reviewed: 2026-10-26 Owner: ax-code runtime

Choose a connection

A provider is the connection to a model service; a model is the specific model selected through it. Connecting a provider does not select every model or grant account access.

In the terminal UI, enter /connect. Choose the path below. Desktop has its own provider settings and bundled runtime; these slash commands describe the terminal UI.

Cloud API: choose API Cloud Provider, select your service and follow the credential prompts. A chat subscription does not necessarily include API access or API credit.
Coding CLI: install and sign in to the supported CLI first, then choose CLI Provider. AX Code reuses that local login; capabilities depend on the adapter.
Local model: start Ollama or LM Studio and load a tool-capable model, then choose Local LLM runtime and confirm the endpoint. AX-Engine runtime has its own setup and model download flow on supported Macs.
Gateway: choose AX Trust for an AX Trust endpoint, API Cloud Provider → Custom API provider for another gateway, or Private GPU cloud for a dedicated compatible deployment. Enter the service URL, token and exact advertised model IDs.

Check the connection

Run these commands in an interactive terminal. The login command prompts for the provider and credentials; do not paste a real key into a shared command or repository file.

ax-code providers login
ax-code providers list
ax-code models

You should see your provider and usable model IDs. A saved credential or successful model listing alone does not prove inference permission. Continue with model selection and send a small test request.

Manage credentials and configuration

Use /providers to manage connections and /connect to replace a key. Disable retains saved credentials; disconnect removes the stored credential. Environment variables and configured keys can still supply access.

Use project ax-code.json for shared non-secret settings, or global ~/.config/ax-code/ax-code.json for your own defaults. Keep secrets out of version control; use the provider credential prompt or environment references described in the gateway guide. Restart after manual configuration edits.

If setup fails

Authentication failure: confirm the key belongs to the chosen service, endpoint and API plan. For a coding CLI, check its own login first.
Empty model list or connection refused: confirm the local server is running, a model is loaded and the URL/port is correct. Ollama normally uses http://localhost:11434/v1; LM Studio uses http://localhost:1234/v1.
Model rejected: use the exact ID from the endpoint and check account permissions. For gateways, a display name may differ from the upstream request ID.
Missing provider: check enabled_providers and disabled_providers. For installation diagnostics, run ax-code doctor; this does not validate your provider account.

Next steps

Provider catalog and login methods · Custom gateways and private GPU endpoints · Choose a model · Create an AX Wiki · Documentation hub