Open source. Desktop first.

Keep the work.
Keep control.

A desktop home for reliable AI development. Explore the early preview: create workspaces, run scripted tasks, and return to a durable record of what happened.

Build the macOS preview

Apple Silicon v0.0.1 source build

Signed installer coming later.

One place to follow the work

A real app. A visible record.

Desktop prototype
Argent desktop prototype showing a local workspace, task history, and verification details for a scripted run.
The installed Argent app, running the local simulation. Preview content is illustrative task input; the recorded state and verification come from the engine.
01 / PERSISTENCE

Come back to your work.

Workspaces, tasks, and activity live in a local database. Reopen the app and inspect the same history.

02 / VISIBILITY

See what actually happened.

Follow execution events and inspect the saved artifact with its independent verification result.

03 / CONTROL

Uncertainty stays visible.

Cancel work explicitly. Interrupted runs require review instead of silently repeating an uncertain action.

The desktop preview

Install locally.
Explore at your pace.

No cloud account or model credentials are needed for this local simulator. The execution service runs on your Mac; local work depends on it staying awake.

Read the project on GitHub
  1. Build from source

    Follow the repository's macOS build instructions to create the local Apple Silicon app. A signed public download is not available yet.

  2. Open your local build

    Open the generated Argent.app from the release directory. The packaged app includes its runtime.

  3. Launch Argent

    Create a local workspace and submit a scripted task. Reopen the app to inspect the persisted result.

Development build. This local preview is unsigned and has not been notarized. macOS may prevent it from opening. A signed public release, automatic updates, and Windows/Linux installers remain future work.

Built in the open

The foundation before the swarm.

Argent’s direction is persistent local and cloud environments, official coding-agent integrations, and changes backed by test evidence. This preview starts with the durable execution model that experience needs.

Follow the project