How to Use TokenAir with Continue: Chat and Agent Limits
TokenAir can significantly lower Continue model spend when suitable chat and editing tasks use a lower-cost model that still meets your review bar. Continue supports an OpenAI-compatible apiBase. Agent mode adds a tool-use requirement that needs a separate test for the selected TokenAir model.
Compatibility verdict
Limited setup
Continue can route chat, edit, and apply roles to TokenAir through provider: openai and apiBase. Agent mode remains limited until a native tool call completes the full round trip.
A custom endpoint path exists, but important features or request shapes need separate testing.
Before you start
- A current Continue installation using config.yaml.
- A TokenAir API key and exact account-enabled model ID.
- A plan to test chat and edit before declaring the model capable of Agent mode.
Setup and compatibility steps
Step 1
Open the Continue configuration
Edit the config.yaml used by your local Continue agent. Keep the API key in a supported secret or local configuration that is not committed.
Step 2
Add the TokenAir model
Use provider: openai, the TokenAir apiBase, and an exact model ID. Set useResponsesApi to false so Continue uses TokenAir's documented Chat Completions surface.
Step 3
Verify chat and edit roles
Select the new model and try a short chat plus one bounded edit. Confirm the output before adding Agent-specific capabilities.
Step 4
Gate Agent mode on a tool test
Add tool_use to the model capabilities after a real native tool call, tool result, and final answer succeed through TokenAir.
Example Continue config.yaml
name: TokenAir
version: 0.0.1
schema: v1
models:
- name: TokenAir coding model
provider: openai
model: YOUR_TOKENAIR_MODEL_ID
apiBase: https://api.tokenair.ai/v1
apiKey: YOUR_TOKENAIR_API_KEY
useResponsesApi: false
roles:
- chat
- edit
- applyReplace placeholders with values from your own TokenAir account. Never commit an API key to source control.
How to verify it
- Continue loads the configuration and displays the TokenAir model.
- Chat and a bounded edit complete through Chat Completions.
- If Agent mode is enabled, a native tool call returns valid arguments and a tool result.
- The final result passes review with retries included in the cost calculation.
How to lower cost without hiding quality loss
TokenAir gives the client access to premium and lower-cost model choices through one OpenAI-compatible API. Savings depend on matching each task with a model that still passes review.
- Start with chat, edit, and apply roles that have clear review points.
- Do not mark a lower-cost model as tool_use capable until the complete tool loop works.
- Use separate models by role when Agent reliability needs a stronger route than routine editing.
Limits to know before production
- Continue can default some OpenAI model families to the Responses API, which TokenAir's public docs do not currently cover.
- Agent mode requires tool_use. Setting the capability flag does not add tool support to a model or endpoint.
- Autocomplete can use different request patterns and should not be inferred from chat or edit success.
FAQ
Can Continue use TokenAir through apiBase?
Yes. Continue documents provider: openai with a custom apiBase and API key for OpenAI-compatible providers.
Why set useResponsesApi to false?
Continue can use /responses for some model families. Setting this value to false keeps the request on TokenAir's documented /chat/completions surface.
Should I add capabilities: tool_use immediately?
No. Add it only after the selected model completes a native tool call and tool-result round trip through TokenAir.
Official sources and related TokenAir docs
Tool settings change between releases. We checked these notes on July 17, 2026. Review the linked tool docs again before a production rollout.
Try one workflow before changing the default.
Use the cost checkup to choose a bounded test, then compare cost per accepted result with your current route.