Local-first by design: what stays on your machine

When you point a tool at your source code, the first question should be where that code goes. For Dispatcher, the answer for ordinary use is: nowhere. The desktop app is built so the sensitive parts of agent-driven development stay on your machine.

What stays local

For local desktop use, these remain on your device and are not uploaded to Dispatcher's servers:

• Repository source code and the worktrees agents work in
• Diffs, patches, and review artifacts
• Terminal output and agent transcripts
• Local workspace and agent-run state, stored on disk by the desktop app
• The API keys and logins your agent CLIs use

Dispatcher runs the agent CLIs you already have installed, signed in with your own accounts. When an agent sends code to its model provider, that happens under your agreement with that provider, using credentials you control. Dispatcher is not in the middle of that traffic.

What the cloud is for

Some things are inherently shared. When you sign in and use cloud features, Dispatcher can store:

• Tasks, projects, cycles, and roadmaps on a shared board
• Knowledge documents and Sites you create or sync
• Organization membership and roles
• Tokens for integrations you connect, such as GitHub, Linear, and Slack

Cloud sync is optional. Running agents in isolated worktrees, reviewing diffs, and opening pull requests all work locally. Turning on cloud features is a deliberate choice to share collaboration data with your team, not a prerequisite for using the app.

Why we drew the line here

Two reasons.

Trust. Engineering teams are rightly cautious about tools that read their code. A product whose core loop never uploads source code is far easier to approve than one that promises to handle it carefully.

Fit. The valuable part of an agent orchestrator happens next to the code: creating worktrees, running terminals, reading diffs. Doing that locally is faster and simpler than shipping a repository to a server and back. The shared parts, a board and a knowledge base, are the parts that actually benefit from being in the cloud.

Read the details

The Privacy Policy describes this boundary in full, and the subprocessors page lists the providers behind the optional cloud features. If you have questions about a specific deployment, email privacy@getdispatcher.dev.