Claude Code (CLI + Desktop)

Claude Code speaks the Anthropic Messages protocol. Point it at a lane root (no /v1) and it routes every call through OpenGateway.

Desktop

1. Open Settings → Connection → Gateway.

2. Fill in:

   Connection:         Gateway
   Gateway base URL:   https://api.opengateway.one/oss
   Gateway API key:    YOUR_KEY
   Gateway auth scheme: bearer
   Model list:         leave empty

3. Click Test connection. For Claude frontier models, use https://api.opengateway.one/frontier instead.

Tip

If the picker still shows old models after switching lanes, restart Claude Code Desktop so it clears its model-discovery cache.

CLI

Set the base URL and auth via environment variables (or ~/.claude/settings.json under env):

export ANTHROPIC_BASE_URL="https://api.opengateway.one/oss"
export ANTHROPIC_AUTH_TOKEN="YOUR_KEY"   # → Authorization: Bearer
# or: export ANTHROPIC_API_KEY="YOUR_KEY"  # → x-api-key

Model discovery (optional)

Claude Code only calls the gateway’s GET /v1/models when you opt in:

export CLAUDE_CODE_ENABLE_GATEWAY_MODEL_DISCOVERY=1

The picker filters to ids beginning claude/anthropic, which is why OpenGateway’s curated aliases are named claude-turbo-hub-*. Discovery is optional — manual ids always work through /v1/messages.

Effort

Extended thinking is on by default. Newer models use CLAUDE_CODE_EFFORT_LEVEL (low/medium/high); older models use MAX_THINKING_TOKENS. OpenGateway maps these onto whatever the upstream model expects — see Effort levels.

Models to try

| Lane | Example model | | --- | --- | | /oss | claude-turbo-hub-qwen3-coder, claude-turbo-hub-gpt-oss-120b | | /frontier | turbo-agent-model-claude-opus-4-7, sonnet-4.6 |