MulmoTerminal
MulmoTerminal is an open-source developer tool from receptron that runs multiple AI coding agent sessions in parallel and displays them as a grid of live terminals in the browser. It is started with a single npx command and targets anthropic's Claude Code and OpenAI's Codex command-line agents, letting a developer supervise several concurrent agents without juggling separate terminal windows. Each cell in the browser grid is a real PTY (pseudo-terminal), so the agents behave exactly as they would in a normal shell. Statuses are colour-coded so the user can see at a glance which agent is working, idle, or waiting for input, and sessions are backed by tmux, which keeps them alive and re-attachable if the browser or the process is closed. Every cell is paired with its own git worktree, so parallel agents edit isolated working copies of the repository rather than conflicting in a single checkout. The tool runs locally on the user's machine and is distributed as an MIT-licensed project on GitHub, making it part of the emerging category of local AI-agent orchestrators.
- Category
- AI Agents
- Pricing
- Open Source
Problem solved
Running several AI coding agents at once normally means many terminal tabs, lost context when a terminal closes, and conflicting edits from agents writing to the same working tree. MulmoTerminal solves this by presenting all sessions as an at-a-glance browser grid with colour-coded status, keeping them alive through tmux persistence, and isolating each session in its own git worktree.
Use cases
- Running several Claude Code or Codex agents on different tasks in parallel
- Monitoring which agent is waiting for input via colour-coded status cells
- Keeping long-running agent sessions alive across terminal or browser restarts with tmux
- Isolating parallel agents in separate git worktrees to avoid edit conflicts
- Reviewing and comparing output from multiple agent sessions side by side
- Spinning up a multi-agent workspace locally with a single npx command
Audience
- Developers
- ML Engineers
- Hobbyists
- Startups
Pros
- One-command startup via npx
- Live PTY sessions in a browser grid with colour-coded status
- tmux-backed persistence keeps sessions alive and re-attachable
- Per-cell git worktree isolates parallel agent edits
- Open source under the MIT license and runs locally
Cons
- Requires tmux and a Node/npx environment to be installed locally
- Targets Claude Code and Codex agents specifically
- Browser grid is aimed at local single-machine use rather than remote teams
- Early-stage project with a small community relative to established terminal multiplexers
Official website: https://github.com/receptron/mulmoterminal
Canonical LambdaBase page: https://www.lambdabase.com/tools/mulmoterminal