Does ATLAS do any direct LLM requests? Or what is the OpenRouter API Key and default model used for?

[GitMurf]

I am wondering if any parts of the ATLAS MCP server reaches out to an LLM directly as I noticed in the README an environment variable for open router api key and default model. Curious what these are for? Does the deep research tool directly make requests to LLM or something?

The main reason I am asking is that I have a bunch of LLM api keys in my environment variables for other stuff, but I do not want ATLAS using these if I don't know about it. I typically use my GitHub Copilot subscription for LLM communication since it is "free" with my subscription, so if there is other LLM communication going on from this MCP server I need to know about it to make sure I am not being charged from one of my other API Keys / LLM services.

# OpenRouter LLM Configuration (Optional - additional parameters present in `src/config/index.ts`)
# OPENROUTER_API_KEY=your_openrouter_api_key # API key for OpenRouter services.
LLM_DEFAULT_MODEL="google/gemini-2.5-flash-preview:thinking" # Default LLM model.

Thanks in advance!! This MCP server is amazing!

[cyanheads]

Hi @GitMurf
Thanks for the kind words!
No, Atlas does not currently utilize the LLM factory or OpenRouter implementation. That was a holdover from my template repo and some experimenting I did but no plans to integrate LLM calls directly in Atlas right now.

I've removed those files in v2.8.15

Let me know if you have any questions!