Skip to content
AgentMux

Quickstart

Get AgentMux running with your first agent in under 5 minutes.

1. Install AgentMux

Section titled “1. Install AgentMux”

Download from agentmux.ai for your platform:

  • macOS — .dmg installer (Apple Silicon)
  • Windows — .exe installer or portable .zip
  • Linux — AppImage

See Installation for detailed steps.

2. Launch and Orient

Section titled “2. Launch and Orient”

When AgentMux opens, you’ll see a default terminal pane. This is the agent operating environment — agents run here as first-class panes, not terminal sessions bolted on afterward. The key areas are:

  • Top bar — Tab management and widget launcher (right side icons)
  • Workspace — Your pane grid, drag to rearrange
  • Status bar — Connection status and system info

If any provider CLIs are missing, open the Toolchain Manager (hamburger menu ≡ → Toolchain Manager) — it shows the detected version, path, and status of every dependency, with install links.

3. Create a Memory bundle

Section titled “3. Create a Memory bundle”

Open an agent pane (Cmd+Shift+A / Alt+Shift+A, or click the Agent icon in the top bar), then open the agent pane’s settings panel (cog icon in the pane header) and switch to the Memory tab. See Memory bundles for the full configuration surface.

Click + New Memory and fill in:

FieldExample
Namemy-claude
ProviderClaude Code
Working Directory/home/user/projects/myapp

Click Create.

4. Launch an Agent Session

Section titled “4. Launch an Agent Session”

In the Launch Agent modal, pick your Memory bundle (and optionally an Identity bundle for credentials). Click Launch to open an agent pane.

The first time you pick a provider, AgentMux installs that CLI for you — an inline install modal runs npm install and streams the output. No npm install -g step beforehand. Once installed, the binary is cached per AgentMux version and reused on every later launch.

The agent starts in a new pane within your workspace. You’ll see:

  • Real-time streaming output
  • Tool calls as they execute
  • File diffs when the agent writes files

5. Add More Panes

Section titled “5. Add More Panes”

Run a terminal or a second agent alongside your first:

  1. Press Cmd+D (Alt+D) to split right — the new pane opens as a terminal
  2. To launch a second agent, press Cmd+Shift+A (Alt+Shift+A) to open the agent picker
  3. Navigate between panes with Ctrl+Shift+Arrow

Each pane is independent — agents don’t share context unless you wire them explicitly via the MuxBus or Agent App API.

6. Monitor System Resources

Section titled “6. Monitor System Resources”

Click the chart icon in the top bar to add a Sysinfo pane. It shows live CPU, memory, network, and disk I/O graphs.

7. Add More Agents

Section titled “7. Add More Agents”

AgentMux supports multiple agents running simultaneously. Open the Memory pane again and create bundles for different providers:

  • Claude Code — claude -p --output-format stream-json --verbose --include-partial-messages --dangerously-skip-permissions
  • Codex CLI — codex exec --json --dangerously-bypass-approvals-and-sandbox -
  • Gemini CLI — gemini --output-format stream-json --yolo -p ""
  • OpenClaw — acpx --agent openclaw
  • Kimi Code CLI — kimi --print --output-format stream-json --yolo -p ""
  • GitHub Copilot CLI — copilot --acp
  • Pi — pi --json

Each agent runs in its own pane. Use the Swarm view to monitor all of them at once.

Next Steps

Section titled “Next Steps”