Agent Manager
Open-source tool that runs and manages multiple AI coding agents as tmux sessions in the terminal
About Agent Manager
Agent Manager is a terminal UI tool written in Go that runs each AI coding agent as a live tmux session. Every row in the list is a real session running a coding CLI, color-coded by panel state, so you can read, reply, and review all agents from one screen. Press space on a group row, type a task, and hit enter — the agent is already running with your prompt and group directory, no forms, cd, or naming needed. It has built-in status detection for 8 coding CLIs including Claude Code; other CLIs connect with a short TOML snippet. It supports macOS, Linux, and Windows via WSL2, requires tmux 3.1 and git, and is free and open source under Apache-2.0.
Best for
Best use cases
Press space, type one sentence, and instantly launch an agent for a new task
Watch the live status of each coding CLI in a single-screen list and reply one by one
Connect your own coding CLI to unified management with a short TOML snippet
Features
Live tmux session list
Every row is a tmux session actually running a coding CLI, color-coded by panel state.
One-sentence task dispatch
Press space on a group row, type the task, and hit enter — the agent is already running with your prompt and group directory.
Multi-CLI status detection
Claude Code, Codex, OpenCode, Grok, Gemini, Pi, Command Code, and Hermes work out of the box; other CLIs connect with a short TOML snippet.
Cross-platform support
Supports macOS and Linux on amd64 and arm64, plus Windows via WSL2.
Pros
- Free and open source under Apache-2.0, with no accounts, license keys, or paid tiers
- Launch an agent session with prompt and group directory from a single sentence
- Built-in status detection for 8 coding CLIs; other CLIs connect with one TOML snippet
- Covers macOS, Linux, and Windows under WSL2
Cons
- The official page notes that cost tracking is not yet implemented
- Mouse-driven list navigation is not yet implemented