Skip to main content
Continue reads its model settings from a YAML config, where each model block sets provider, model, and apiBase. Point that apiBase at https://api.tokios.com/v1 with an sk-tok-… key and your Tokios deployment name as the model, and Continue reaches your local model over one authenticated endpoint — no open port on your LAN required.

​
Prerequisites

  • A running Tokios connector paired with your local model — see Connector install
  • A registered model deployment name — see Register a model
  • A Tokios API key (sk-tok-…) — see API Keys
  • Continue installed in your editor

​
Configure Continue

1

Open Continue's config file

Continue reads a YAML config from your home directory:
The directory is .continue, with a leading dot. Older Continue releases used config.json instead — if that’s what you have, either migrate to YAML or translate the block below into the JSON schema your version expects.
2

Add an OpenAI-compatible model block

Add a model entry using an OpenAI-compatible provider, with apiBase pointed at Tokios and model set to your public deployment name — for example, gemma-tunnel:
model is the deployment name you registered in the Models tab, not your Ollama, llama.cpp, vLLM, or LM Studio local model id. See Continue’s configuration docs for the full set of config keys.
3

Point chat, edit, and autocomplete at the same model

Continue lets you assign different roles — chat, edit, autocomplete — to different models. Reference the same Tokios model block for each role you want served by your local model. Add the model to the tabAutocompleteModel and chat sections of your config:
Each section is an independent model list — the models block at the top registers the model, and the role sections assign it. See Continue’s configuration docs for the full set of role-assignment keys.
Replace gemma-tunnel with the deployment name you registered, and sk-tok-YOUR_KEY with your own key. Never commit either into a shared or checked-in config file.

​
Don’t bind Ollama to 0.0.0.0

Continue’s documented way to reach an Ollama instance on another machine is to set OLLAMA_HOST=0.0.0.0 on the machine running Ollama and point apiBase at that machine’s raw LAN IP (for example http://192.168.1.50:11434). That makes Ollama listen on every network interface, and Ollama ships with no built-in authentication — anything on the network that can reach the port can list models and run inference with no credential. Keep Ollama on 127.0.0.1 and run the Tokios connector next to it instead. The connector dials out to https://api.tokios.com; Ollama never opens an inbound port, and every request to it needs a valid sk-tok-… key:
  • No unauthenticated exposure — nothing needs to reach your LAN or your machine’s port directly.
  • Same apiBase from any network — https://api.tokios.com/v1 works whether Continue is on the same LAN or a different building entirely.
  • Scoped, revocable keys — cut off a leaked key on the Keys tab without touching Ollama or the connector.

​
Troubleshooting

The API key Continue sent is missing, malformed, or has been revoked. Confirm your config’s apiKey holds the full sk-tok-… key with no extra whitespace, and check its status on the Keys tab.
The key exists but isn’t scoped to the deployment you referenced in model, or your account is suspended. Check the key’s model patterns on the Keys tab.
The model value in your config doesn’t match a registered deployment. Confirm it exactly matches the public name on the Models tab — not your Ollama/llama.cpp/vLLM/LM Studio local model id.
The connector for that deployment is offline or at its concurrency cap. Check the Connectors tab — the connector needs to show as Online, and the machine running it needs to still be up and reachable.

​
Next steps

Ollama remote access

See the full case against binding Ollama to 0.0.0.0, port forwarding, and tunnels — and the outbound-only alternative.

Compatible clients

See base URL and key setup for other OpenAI- and Anthropic-compatible clients.