LM Warden speaks the three protocols coding agents use: OpenAI Chat Completions, the Anthropic Messages API and the OpenAI Responses API. So Claude Code, the Anthropic and OpenAI SDKs, OpenCode, Aider, Grok CLI, Continue, Cline, Cursor and Codex CLI connect with a base URL and a key, and their requests show up per key, per session, with the phase each one is in.
Where a request from a coding agent goes. Every agent sets the same two things, a base URL and a key; the warden checks the key, queues the request and sends it to the replica that already holds the session’s cache. Each agent’s page draws its own settings.
In words:Your coding agent sends its requests to LM Warden with base URL set to the warden and API key holding a warden key. LM Warden checks the key, speaks the OpenAI and Anthropic APIs, queues the request by key priority and hands it to the replica router, which keeps each session on the replica that already holds its cache, one model copy per GPU. A separate path goes to api.anthropic.com: Claude Code only: other Claude models go here, on your own login.
You don’t write the config by hand. In the console, Connect picks your client, your model and your key and writes the files to paste. Each page below shows the same files with placeholders and what works.