Get Started

Run Mewbo in your terminal¶

Run Mewbo agent sessions in a full screen terminal app, against your real files and your real shell. Launch it from your shell, type a request, and watch the run stream live on one screen.

You drive an entire run from the keyboard. The engine runs on your machine, so the core loop makes no API round trip.

The web console and the Android app fit a graphical surface or a session on your phone. The terminal fits speed and locality at your desk.

The interface is one Textual application, launched from cli_master.py and defined in app.py.

Prerequisites¶

Requirement	Notes
Python 3.10+	uv manages the virtualenv

The terminal client ships with the base install, so no extra is needed.

Install and first run¶

Install the dependencies with uv sync.
Launch the app with uv run mewbo. It opens full screen.
On a fresh machine, type /init and press Enter to scaffold the config files Mewbo needs.
Type a request in plain language and press Enter.
Approve or deny when the agent asks. Mewbo prompts you before any write or shell command.
Type /help to list the commands, and /exit when you are done.

Run uv tool install . from the repository root to put mewbo on your PATH and start a session from any directory.

Configuration¶

Pass --config at launch to name one specific file. Otherwise Mewbo walks a config chain rooted at your project.

LLM Setup covers provider keys and model choice.
Configuration Reference lists every config key on the site.

Next steps¶

The Interface The transcript, tool cards, the plan dock and the approval modals. Agent Fleet The sidebar tracking every sub-agent, and the drill in view behind each row. Remote Sync Local by default, and what reaches the network when you opt in. Configuration Launch flags, the config chain and session recovery.