Coding agents

Export your SayGM key before starting a terminal-based agent:

export GM_API_KEY="gm_live_…"

Claude Code

Claude Code uses SayGM’s Anthropic-compatible endpoint.

export ANTHROPIC_BASE_URL="https://api.saygm.com"
export ANTHROPIC_AUTH_TOKEN="$GM_API_KEY"

claude --model claude-opus-4-8

PowerShell:

$env:ANTHROPIC_BASE_URL="https://api.saygm.com"
$env:ANTHROPIC_AUTH_TOKEN=$env:GM_API_KEY
claude --model claude-opus-4-8

Codex CLI

export OPENAI_BASE_URL="https://api.saygm.com/v1"
export OPENAI_API_KEY="$GM_API_KEY"

codex --model gpt-5.5

Cursor

Open Settings → Models → OpenAI API Key, enable Override Base URL, and use:

Base URL   https://api.saygm.com/v1
API Key    your SayGM API key
Model      an available chat.completions model ID

Add the model as a custom model if it is not already listed.

Aider

export OPENAI_API_BASE="https://api.saygm.com/v1"
export OPENAI_API_KEY="$GM_API_KEY"

aider --model openai/gpt-5.5

The openai/ prefix belongs to Aider’s provider selection, and Aider strips it before the request reaches SayGM. Put it in front of the whole catalog id, including any slashes the ID already contains — a confidential model becomes --model openai/qwen3-32b-tee. See choose a model for the model-ID rule.

Cline and Roo Code

For an OpenAI-shaped model, configure:

API Provider   OpenAI Compatible
Base URL       https://api.saygm.com/v1
API Key        your SayGM API key
Model ID       an available chat.completions model ID

For a Claude model, choose the Anthropic provider and enable the custom base URL:

API Provider          Anthropic
Use custom base URL   https://api.saygm.com
API Key               your SayGM API key
Model ID              an available messages model ID

Continue

Add an OpenAI-compatible model to ~/.continue/config.json:

{
  "models": [
    {
      "title": "SayGM",
      "provider": "openai",
      "model": "gpt-5.5",
      "apiBase": "https://api.saygm.com/v1",
      "apiKey": "your SayGM API key"
    }
  ]
}

Prefer Continue’s environment-variable or secret support where available instead of storing the key directly in a file.

Zed

Add an OpenAI-compatible model to Zed’s settings.json:

{
  "language_models": {
    "openai": {
      "api_url": "https://api.saygm.com/v1",
      "available_models": [
        { "name": "gpt-5.5", "max_tokens": 200000 }
      ]
    }
  }
}

Then export the key before starting Zed:

export OPENAI_API_KEY="$GM_API_KEY"

Check the selected model’s actual context and output limits rather than treating the example max_tokens value as a capability guarantee.