Skip to content
Lattis Lattis

Introduction

Lattis is a desktop workspace for getting more done with AI coding agents. Instead of babysitting one agent in one terminal, you run a whole fleet — each on its own branch — and Lattis keeps you in command of all of them.

It’s built around three jobs:

  1. Run many agents at once. Launch Sawyer (built in) plus Claude, Codex, Pi, and OpenCode side by side, each in its own git worktree, without them stepping on each other.
  2. Always know what needs you. Every agent and branch carries a status you can read at a glance — who’s working, who’s blocked, who’s waiting on you — and pull requests, reviews, and CI live right in the app.
  3. See what it costs. Every request is attributed and priced, so you know the dollar cost per project, model, and conversation as the work happens.

Run a fleet, not a single agent

Section titled “Run a fleet, not a single agent”

Add a project (a git repo) and open a branch — Lattis spins up an agent in a fresh worktree with its own terminal. Open as many branches as you like and run several agents in parallel, even two on the same task to compare. The bottleneck stops being how many agents you can start, and becomes how fast you can review what they produce.

Never lose track

Section titled “Never lose track”

The hard part of parallel work is knowing where to look next. Lattis answers that with stoplights: a green dot is working, amber is waiting on you, red has errored. Each branch also shows its live pull request and CI state — ready, awaiting review, blocked — so you can review, merge, push, and pull without ever leaving the app.

Ship as a team

Section titled “Ship as a team”

A fleet of agents opens pull requests faster than a person can read them, so Lattis takes on the other side too. Pending Reviews lists every pull request your team asked you to read, and one click opens the branch in a review workspace: the whole branch as one guided diff, draft comment threads on the lines they describe, and an optional read-only agent pass that finds issues for you to triage. Your verdict posts to GitHub as anchored line comments. Asking for a review is the same one click, with a reviewer picker that ranks the people who wrote the code you changed.

Know what every agent costs

Section titled “Know what every agent costs”

Because every agent’s traffic flows through one local endpoint, Lattis attributes and prices it. The usage and cost view breaks spend down by project, model, and conversation, so you can decide what belongs on a frontier model and what a free local model can handle. For team-wide reporting, the optional web app rolls the same numbers up across everyone’s machines.

Where to next

Section titled “Where to next”