Goose

Goose adds UberLLM as a custom provider described by one JSON file.

Before you start

• An UberLLM account with credit, and an API key from Dashboard → Keys (ull_...).

Setup with a file

1. Save this as uberllm.json in Goose's custom_providers folder:
   • macOS / Linux: ~/.config/goose/custom_providers/
   • Windows: %APPDATA%\Block\goose\config\custom_providers\

{
  "name": "uberllm",
  "engine": "openai",
  "display_name": "UberLLM",
  "description": "DeepSeek, GLM, Kimi, Gemini, Claude, Qwen and other models through one OpenAI-compatible API key.",
  "api_key_env": "UBERLLM_API_KEY",
  "base_url": "https://api.uberllm.dev/v1/chat/completions",
  "headers": { "x-title": "goose" },
  "models": [
    { "name": "deepseek/deepseek-v4.1-flash", "context_limit": 1048576 },
    { "name": "z-ai/glm-5.2", "context_limit": 1048576 },
    { "name": "moonshotai/kimi-k3", "context_limit": 1048576 },
    { "name": "google/gemini-2.5-flash", "context_limit": 1048576 },
    { "name": "anthropic/claude-haiku-4.5", "context_limit": 200000 }
  ],
  "supports_streaming": true,
  "requires_auth": true
}

2. Set the key and start a session:

export UBERLLM_API_KEY="ull_your-key"
goose session start --provider uberllm

To keep the key in Goose's keychain instead, open the UberLLM provider in Goose Desktop and enter the key there.

Setup with goose configure

Run goose configure → Custom Providers → Add A Custom Provider, then:

• API Type: OpenAI Compatible
• Name: UberLLM
• API URL: https://api.uberllm.dev/v1/chat/completions
• Authentication Required: Yes → Static API key → paste your key
• Available Models: deepseek/deepseek-v4.1-flash,z-ai/glm-5.2,moonshotai/kimi-k3
• Streaming Support: Yes

Recommended models

Checked on 2026-10-30: each one answered and made a tool call through UberLLM.

Goose may send openai/gpt-5* ids to the OpenAI Responses API. UberLLM serves that route too (Responses API), so those ids work as well.

Common errors