DEV Community

FuturMix
FuturMix

Posted on

How to Use Aider with Any LLM Provider (Custom API Endpoint)

#ai #python #productivity #tutorial

Aider is one of the best AI pair programming tools, but by default it connects directly to OpenAI or Anthropic. What if you want to route through a custom API endpoint — for cost savings, failover, or unified billing?

Here's how to set it up.

The Problem

You're using Aider and want to:

  • Access multiple model providers (Claude, GPT, Gemini) without juggling API keys
  • Route through a gateway for reliability or cost optimization
  • Use a self-hosted or third-party API endpoint

Solution: Environment Variables

Aider uses the OpenAI SDK under the hood, which means it respects OPENAI_API_BASE for custom endpoints.

Step 1: Set Environment Variables 

export OPENAI_API_BASE=https://futurmix.ai/v1
export OPENAI_API_KEY=your-gateway-key
Enter fullscreen mode Exit fullscreen mode

Step 2: Run Aider with Any Model 

# Use Claude Sonnet 4.5
aider --model claude-sonnet-4-5-20250929

# Use GPT-5.4
aider --model gpt-5.4

# Use Gemini 2.5 Pro
aider --model gemini-2.5-pro
Enter fullscreen mode Exit fullscreen mode

That's it. One key, one endpoint, any model.

Step 3: Make It Permanent

Add to your .bashrc or .zshrc:

# AI Gateway config
export OPENAI_API_BASE=https://futurmix.ai/v1
export OPENAI_API_KEY=your-gateway-key
Enter fullscreen mode Exit fullscreen mode

Or create a .env file in your project:

OPENAI_API_BASE=https://futurmix.ai/v1
OPENAI_API_KEY=your-gateway-key
Enter fullscreen mode Exit fullscreen mode

Using .aider.conf.yml

You can also configure this in Aider's config file:

# ~/.aider.conf.yml
openai-api-base: https://futurmix.ai/v1
openai-api-key: your-gateway-key
model: claude-sonnet-4-5-20250929
Enter fullscreen mode Exit fullscreen mode

Why Use a Gateway?

Benefit Direct API Through Gateway
API Keys needed One per provider One total
Model switching Change key + SDK Change model param
Failover Manual Automatic
Billing Per-provider Unified

Troubleshooting

"Model not found" error:
Make sure you're using the full model ID (e.g., claude-sonnet-4-5-20250929, not claude-sonnet).

Authentication errors:
The gateway should accept Bearer tokens in the standard Authorization header. Check that OPENAI_API_KEY is set correctly.

Slow responses:
Gateway routing adds minimal latency (typically < 50ms). If you're seeing significant delays, check your network connection to the gateway endpoint.

Other Tools That Work the Same Way

This same OPENAI_API_BASE pattern works with:

  • Cursor — Settings → Models → OpenAI API Base
  • Continue — config.json → provider base URL
  • Roo Code — OpenAI Compatible provider settings
  • Claude Code — ANTHROPIC_BASE_URL environment variable

Any tool built on the OpenAI SDK supports custom base URLs.

Using a custom API endpoint with Aider? Let me know your setup in the comments.

Top comments (0)