How to Use TokenAir with Zed Agent: Capability Limits
TokenAir can significantly lower Zed Agent model spend when suitable coding tasks use a lower-cost model that still completes tools correctly. Zed supports custom OpenAI-compatible providers. Check its default capability assumptions against the selected TokenAir model.
Compatibility verdict
Limited setup
Zed accepts an API URL, key, model ID, context window, and capability settings for a custom provider. Tool behavior through TokenAir remains model-specific and unverified.
A custom endpoint path exists, but important features or request shapes need separate testing.
Before you start
- A current Zed release with Add Provider under Agent Settings.
- A TokenAir API key, exact model ID, and verified context window.
- A small Agent task that exercises one file tool without risky changes.
Setup and compatibility steps
Step 1
Open Zed Agent Settings
Run agent: open settings, find LLM Providers, and choose Add Provider for an OpenAI-compatible endpoint.
Step 2
Add the TokenAir provider
Enter TokenAir as the provider name, add the API URL, model ID, and the model's verified context window. Store the key through Zed's provider settings so it stays out of settings.json.
Step 3
Keep Chat Completions enabled
TokenAir publicly documents Chat Completions, so keep that capability enabled. Do not switch the provider to Responses-only behavior.
Step 4
Test the assumed tool capability
Zed assumes tools are available for compatible models by default. Run a file-read task and confirm the full tool loop before relying on that default.
Zed provider fields to verify
Provider name: TokenAir
API URL: https://api.tokenair.ai/v1
API Key: YOUR_TOKENAIR_API_KEY
Model ID: YOUR_TOKENAIR_MODEL_ID
Context window: YOUR_VERIFIED_CONTEXT_WINDOW
Chat Completions: enabledReplace placeholders with values from your own TokenAir account. Never commit an API key to source control.
How to verify it
- Zed stores the key and shows the TokenAir provider and model.
- A normal Agent panel message completes through Chat Completions.
- The model makes a valid file tool call and consumes the returned result.
- The final response and any proposed edit meet the task's review bar.
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.
- Use a verified context window instead of an oversized placeholder that invites unnecessary input.
- Disable unsupported capabilities and avoid paying for repeated failures caused by a false tool assumption.
- Compare cost per accepted Zed task and keep a stronger route for complex Agent work.
Limits to know before production
- Zed's OpenAI-compatible defaults assume tools are supported, but TokenAir does not yet publish that guarantee for every model.
- The context window must come from verified model information rather than a copied example.
- Responses-only, reasoning, image, parallel-tool, and interleaved-reasoning settings need separate endpoint support.
FAQ
Can Zed Agent use a custom TokenAir API URL?
Yes for endpoint setup. Zed documents an OpenAI-compatible provider that accepts an API URL, key, model ID, and context window.
Why is a context window required?
Zed uses it to plan how much context the model can accept. Enter the selected model's verified value, not the number from an unrelated example.
Does Zed's tools default prove TokenAir tool support?
No. It is a client-side assumption. Verify a native tool call and tool result with the exact TokenAir model before Agent use.
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.