LM Studio

LM Studio provides a polished GUI plus a CLI (lms) and an OpenAI-compatible local server.

Setup

  1. Run openagent --setup.
  2. Choose Local → LM Studio.
  3. OpenAgent installs LM Studio via brew install --cask lm-studio and bootstraps the lms CLI.
  4. Pick a model from the curated catalog.

Curated models

These are LM Studio catalog IDs (passed to lms get):

Model Size RAM
google/gemma-4-e2b 1.5 GB 5 GB
google/gemma-4-e4b 3.0 GB 5 GB
google/gemma-4-26b-a4b 15 GB 18 GB
google/gemma-4-31b 18 GB 20 GB
qwen/qwen3.5-9b 5.4 GB 16 GB
qwen/qwen3.5-35b-a3b 20 GB 24 GB
openai/gpt-oss-20b 12 GB 24 GB

Custom models

Paste any LM Studio catalog ID in the Custom option:

google/gemma-4-e2b@q4_k_m
qwen/qwen3.5-9b
openai/gpt-oss-20b

Server config

LM Studio’s server runs on localhost:1234. OpenAgent talks to its OpenAI-compatible /v1/chat/completions endpoint. To point at a different host, pick Custom → http://… in the picker.

Troubleshooting

“Failed to resolve artifact” — the model name isn’t in LM Studio’s catalog. Use lms get (no args) to see the full staff-pick list and pick a valid publisher/model ID.