How to Set Up Claude Code with a Custom API Endpoint
- Syed Dastgir

- Jul 4
- 1 min read
Claude Code ships with Anthropic's API as the default endpoint. But it supports a custom base_url — pointing it at any OpenAI-compatible API including multi-model routers, local inference servers, or enterprise gateways. This guide shows you how in under 2 minutes.
Why change the base_url?
Access multiple AI models (GPT, Gemini, DeepSeek, local Ollama) without switching tools
Route simple tasks to cheaper or free local models automatically
Get full visibility into what model ran, why it was selected, and what each request cost
Step 1 — Get your API key
Sign up at ai.adorbistech.com — 1,000 free credits, no card required. The key works with any OpenAI-compatible tool.
Step 2 — Set the environment variables
Claude Code reads ANTHROPIC_BASE_URL and ANTHROPIC_API_KEY:
# macOS / Linux
export ANTHROPIC_BASE_URL=https://api.adorbistech.com/v1
export ANTHROPIC_API_KEY=your_api_key_here
claudeOn Windows PowerShell:
$env:ANTHROPIC_BASE_URL = "https://api.adorbistech.com/v1"
$env:ANTHROPIC_API_KEY = "your_api_key_here"
claudeStep 3 — Verify the connection
claude -p "say hello and tell me which model you are"The response comes back from the custom endpoint. A multi-model router will indicate which model was automatically selected.
Step 4 — Use routing-aware model strings
adorbis/auto — fully automatic, routes based on request complexity
adorbis/code — optimized for coding and debugging tasks
adorbis/quick — fastest response for simple tasks
adorbis/reasoning — high-reasoning models for complex analysis
Common issues
401 Unauthorized — Check ANTHROPIC_API_KEY value
Connection refused — Check base_url includes /v1 with no trailing slash
Model not found — Check endpoint documentation for supported model strings
That is the entire setup. For 41 AI models with automatic routing and a full decision record for every request, Adorbis is free to start — no card required.



Comments