Skip to content

Setting up Zed

Zed Agent can call Nexus over OpenAI Chat or Anthropic Messages. External Agents and Terminal Threads keep their own configuration and do not use these fields.

The shared order is on Client setup.

Values

Protocol Base URL Route the client builds
OpenAI Chat https://api.nexus-hub.tech/v1 POST /v1/chat/completions
OpenAI Responses https://api.nexus-hub.tech/v1 POST /v1/responses
Anthropic Messages https://api.nexus-hub.tech POST /v1/messages

OpenAI-compatible steps

  1. Query GET /v1/models and copy an id this key can use on Chat.
  2. Open agent: open settings, then LLM Providers → Add Provider.
  3. Choose the OpenAI-compatible entry. Set the API URL to https://api.nexus-hub.tech/v1, the model id, and the context window from Models and prices.
  4. Enter the key in the UI keychain, or use the variable Zed creates for the entry name. Do not put the key in settings.json.

The same URL with Chat turned off is the Responses case. Use it only for an id that Nexus allows on Responses.

Config file

This fragment has no key. Set max_tokens to 1050000: that is the declared window for every model currently available in Nexus. For a Responses-only model, set chat_completions to false.

{
  "language_models": {
    "openai_compatible": {
      "nexus": {
        "api_url": "https://api.nexus-hub.tech/v1",
        "available_models": [
          {
            "name": "MODEL_FROM_V1_MODELS",
            "display_name": "MODEL_FROM_V1_MODELS",
            "max_tokens": 1050000,
            "capabilities": {
              "tools": true,
              "chat_completions": true
            }
          }
        ]
      }
    }
  }
}

Anthropic Messages

In LLM Providers, add a custom Anthropic endpoint with URL https://api.nexus-hub.tech and a model from GET /v1/models. Enter the key in the UI or in an environment variable. Do not add it to the settings file. Do not append /v1/messages.

Check

Select the model in the Agent Panel and send a short message. A normal reply means the connection works. Streaming depends on the protocol and the model.

If it does not work

  • 401 or invalid_key: check the keychain or the environment variable.
  • model_not_found or model_not_allowed: query GET /v1/models again and use the exact id.
  • unsupported_model: the id is not allowed on this protocol. Switch the entry or the id.
  • 404: compare the API URL with the table. Do not paste the full operation path into the URL field.

Send support the route, error code, time, x-request-id, and the Zed version. Do not send the key or a full prompt.

Official reference: Zed Use API Access.