Zed

Zed  supports custom OpenAI-compatible providers. The current configuration uses language_models.openai_compatible; do not copy older provider formats without checking your installed version.

1. Add a provider

Open the command palette and run agent: open settings. In LLM Providers, choose Add Provider and enter a custom OpenAI-compatible provider:

2. Settings-file alternative

Merge this section into your settings rather than replacing the whole file:

{
  "language_models": {
    "openai_compatible": {
      "fenno": {
        "api_url": "https://api.fenno.ai/v1",
        "available_models": [
          {
            "name": "<MODEL_ID>",
            "display_name": "Fenno model",
            "max_tokens": 8192
          }
        ]
      }
    }
  }
}

Replace <MODEL_ID>. 8192 is an example context budget for Zed, not a statement of the model’s actual limit or an output-token setting. Adjust it to the chosen model and workload.

3. Store the API key

Enter your key in the provider settings UI; Zed stores it in the system keychain rather than settings.json. Alternatively, provider ID fenno maps to FENNO_API_KEY in the process environment. A non-empty environment variable can override a keychain value.

A GUI application already running may not inherit a new shell export. Prefer the settings UI, or launch/restart Zed from the intended environment. Never add an api_key field with a secret to shared settings.

4. Verify the selected model

Select the Fenno provider/model in the Agent panel and start a new conversation with “Reply with OK only.” Check the actual account usage. Then test one safe tool call in a test project, leaving normal approval controls enabled.

Capabilities such as tools and images must match the selected model. An option in Zed’s UI cannot create an API capability. Zed-hosted models, subscriptions and other editor features are separate from your Fenno usage.

For failures, check the selected provider, inherited credentials, exact ID and final /v1/chat/completions path. See Zed’s official API-access guide  and tool troubleshooting.