Connect a client#
Start a model first. OpenAI-compatible clients use
http://127.0.0.1:1920/v1; query /v1/models for the served model ID. A model
served from Modal uses the same URL.
For coding agents, the safer starting point is claude-sandbox, which runs the client in a devcontainer and reduces its access to your host environment. See Run Pi in claude-sandbox for setup. Keep lllm2 running on the host and run the client inside the devcontainer.
If their CLIs are installed, these wrappers configure a local session and forward additional arguments:
lllm2 claude
lllm2 codex
The Claude and Codex wrappers read the running model’s context and slots. Relaunch the client
after changing server context. Use lllm2 claude -- --help (or the equivalent
wrapper) to read the client’s own help.
For Pi, use claude-sandbox’s launcher, which runs it in the same sandbox and discovers the running lllm2 model on each launch:
uv tool install claude-sandbox && claude-sandbox pi