LlaMa Commit Studio
VS CODE EXTENSION·LM STUDIO · OLLAMA·V0.8.5

Commit messages, written on your own machine.

LlaMa Commit Studio reads your staged diff, asks a model you are already running — LM Studio or Ollama — and drops the message in the commit box, ready to edit. No API keys. No cloud. No code leaving your computer.

Get it on the Marketplace See how it works →

FREE · MIT · NO API KEYS · NO TELEMETRY

01 — QUICKSTART

Three steps, and none of them is signing up.

If you already run a local model, you are one click from a commit message. If you do not, step one takes about five minutes.

Three steps: run a model in LM Studio or Ollama, connect from the main screen, then generate the commit message from the Source Control toolbar

THE MESSAGE LANDS IN THE COMMIT BOX, READY TO EDIT — NOTHING IS COMMITTED FOR YOU

A setup walkthrough also opens the first time the extension activates, and can be reopened whenever you need it with LlaMa Commit Studio: Open Setup Guide.

02 — WHAT IT ADDS

Two buttons and a llama.

It lives where you already commit, rather than asking you to work somewhere else.

Sparkle — generate

Reads the staged diff through VS Code's own Git extension, sends it to your local server, and writes the result into the commit box.

Gear — the main screen

A visual panel for the connection and the model, with your past messages listed underneath on a commit-graph rail.

Activity Bar view

Shows the provider and model in use at a glance, with one-click access to the main screen, a generation, server detection and the setup guide.

03 — WHAT IS INSIDE

Built for the parts of local AI that usually go wrong.

A — DETECTION

Finds your server for you

Detection reads LM Studio's own saved port and OLLAMA_HOST before falling back to the defaults, so a non-default port is not a dead end.

B — PROBING

Knows which product answered

LM Studio replies 200 to endpoints it does not have, so a status code proves nothing. A server counts as Ollama only when /api/tags returns a models array — the shape of the reply decides, not the status.

C — MODELS

A list that tells you something

Load state, quantisation, parameter size, context length and vision support, loaded models first. Embedding models are left out and counted underneath — they cannot write text, so offering them only fails later.

D — STYLE

Conventional or freeform

feat(auth): add token refresh, or a plain imperative summary. Reasoning output, markdown fences, wrapping quotes and "Commit message:" labels are stripped before the text reaches you.

E — HISTORY

The last forty messages

Past generations are listed in the clear on a commit-graph rail under the header, kept on your machine like everything else here.

F — PRIVACY

One address, and no other

Requests go only to the base URL you configure. There is no telemetry and no other network access — the extension has nowhere else to send anything.

04 — COMPARE

The same job, without the round trip.

Writing a commit message is a small task. It does not need a subscription, an API key, or your employer's source code on someone else's hardware.

TASK
A CLOUD COMMIT-MESSAGE TOOL
LLAMA COMMIT STUDIO
Where your diff goes
Uploaded to a vendor's servers
To the base URL you configured, and nowhere else
What it costs
A subscription, or billing per token
Nothing — MIT, no account, no key
Working offline
Nothing works
Works on a plane, the model is already on the disk
Setting it up
Sign up, create an API key, paste it in
Point it at a server you already run — the port is detected
Choosing the model
Whatever the vendor ships this month
Anything you have pulled, with its size and context length listed
05 — INSTALL

Install it

One click from the Visual Studio Marketplace, and it keeps itself up to date. Or search LlaMa Commit Studio in the Extensions panel.

Get it on the Marketplace

Or from a terminal:

$ code --install-extension MadeByAriel.llama-commit-studio
WHAT IT NEEDS

VS Code 1.85 or newer with the built-in Git extension enabled, and a local server: LM Studio or Ollama.

06 — SUPPORT

One designer, end to end.

LlaMa Commit Studio is built and maintained by one person and given away free, under MIT. A coffee pays for the test hardware and the next release.

Buy me a coffee →

BUYMEACOFFEE.COM/MADEBYARIEL