Zed Agent with Nbility

Zed’s Agent Panel accepts OpenAI-compatible providers. This configures model conversations, not the separate Edit Prediction provider.

Add the providerCopy link

Open user settings from Zed’s command palette and merge:

{
  "language_models": {
    "openai_compatible": {
      "nbility": {
        "api_url": "https://api.nbility.ai/v1",
        "available_models": [
          {
            "name": "YOUR_MODEL_ID",
            "display_name": "Nbility model",
            "max_tokens": 8192,
            "max_output_tokens": 1024,
            "capabilities": {
              "tools": true,
              "images": false,
              "chat_completions": true
            }
          }
        ]
      }
    }
  }
}

Use the ID and limits for the actual model. max_tokens means context window here; the example limits are illustrative. Match tool and image capabilities to the model.

Store the keyCopy link

Prefer Zed’s provider settings UI for credential storage. Alternatively, set the generated NBILITY_API_KEY environment variable and launch Zed from that terminal. Do not put API keys in settings.json.

Select and verifyCopy link

Open the Agent Panel, select the Nbility model, test plain text, and check Usage logs before tools.

• Missing model: check JSON, reload settings, and confirm it is under openai_compatible.nbility.
• Chat fails with a valid key: match chat_completions to the model’s protocol.
• Missing key after desktop launch: use provider credential settings or check environment inheritance.

ReferencesCopy link

Adapted and rewritten for Nbility from aicentos’s Zed Agent guide, licensed under CC BY 4.0. Its account, plan, and proprietary service instructions do not apply here. See the client’s official documentation.