on-device only

Your work log never leaves your laptop.

Chief tracks what is waiting on you, writes your work log as you go, and answers questions about your own week — on a model that runs on your machine.

Download What leaves the machine Changelog

Version 0.8.0 · no account, no telemetry

0 bytes to a server of ours

1 port on localhost

$0 per token, ever

Chief

briefs

Today

26 Sep

Friday

25 Sep

Thursday

24 Sep

Wednesday

23 Sep

Today

on-device · Llama 3.2 3B · synced 4m ago

Saturday, 26 September

calendar review requests Linear

briefs/2026-09-26.md

Write it again

10:30 design review with Priya and Ravi.
Priya asked for your review on chief.ai#52 two days ago; it blocks her release.
CHF-18, the calendar import, is assigned to you and due Friday.

drafted for you · nothing sent

Ask for a review on chief.ai#51

Not sent

Hi Sam — could you take a look at #51 when you have a moment? It is a one-line fix to the caption bar drag.

Three things it does without being asked

Tells you what to do first

One recommendation at the top of the day with the reason attached. A review that has blocked two people beats an empty inbox.

Writes your work log

Pull requests you merge are summarised into a sentence each as they happen, so standups, one-to-ones and reviews are already written when you need them.

Answers about your own week

“What did I ship?” “What is waiting on me?” Grounded in your log, with the sources listed, answered by a model on your machine.

why local matters here

Your work log is the most sensitive document you own.

It is every unshipped feature, every slipped date and every awkward thread, in one place. Chief keeps it in a SQLite file in your own app directory and asks a model on the same machine to read it. There is no account, no sync, and no vendor with a copy.

Read how it works

Your questions and answers never leave

Your work log database local file

Your notes and drafts markdown you own

Usage analytics none collected

Requests to your tools outbound, your token

Running in four minutes

Install the app

Builds for macOS, Windows and Linux. The llama.cpp engine is bundled, so there is no separate runtime to install and no installer account.

Download a model, once

First run fetches Llama 3.2 3B — about 2 GB, with a progress bar and a resumable download. After that it is yours offline.

1245 of 2019 MB

Connect one tool

GitHub first, via device code in your browser. The token is stored on this machine only.

B7C4-9QLD

Get Chief

Version 0.8.0. No account, no telemetry, no card.

macOS Apple silicon · .dmg macOS Intel · .dmg Windows x64 · .exe Linux x64 · .AppImage Linux x64 · .deb Linux x64 · .rpm