Agent Mux

🚀 Agent Mux

Agent Mux is a tmux-backed agent pool combined with a small MCP (Model Context Protocol) server. It addresses the need for efficient management of interactive agents by leveraging tmux's session and window management capabilities. This allows for seamless creation, listing, switching, and closing of projects, as well as spawning, listing, attaching, interrupting, and killing agents.

🚀 Quick Start

Manual tmux

After creating a project and spawning agents, you can inspect the tmux session:

tmux -L ccx attach -t <project>
# Detach without closing: Ctrl-b d
# Switch windows: Ctrl-b 0/1/2/...

MCP Server

Run the MCP server (stdio):

node servers/agent-mux-server.mjs

Then register it in your MCP client as a stdio server. Example (client formats vary):

{
  "mcpServers": {
    "agent-mux": {
      "command": "node",
      "args": ["servers/agent-mux-server.mjs"]
    }
  }
}

✨ Features

• Dedicated tmux socket: Uses tmux -L ccx ... to avoid interfering with your default tmux.
• One project = one session, one agent = one window (single pane).
• Persistent registry under .ccx/projects/<project>/agents.json.
• Per-agent log file via pipe-pane to .ccx/projects/<project>/logs/<agent>.log.
• Conservative name validation for projects and agents.

📦 Installation

From the repo root:

npm install

💻 Usage Examples

Via MCP tools

Typical flow:

1. Create/select a project
   • create_project({ "name": "my-project", "cwd": "/path/to/workdir" })
   • switch_project({ "name": "my-project" })
2. Spawn an agent
   • spawn_agent({ "type": "codex", "name": "codex-1" })
3. Talk to it
   • send_to_agent({ "name": "codex-1", "text": "hello" })
4. Read output
   • read_agent_output({ "name": "codex-1", "lines": 50 })

As a library (Node.js)

import { AgentMux } from "./lib/agent-mux/index.mjs";

const mux = new AgentMux({ projectRoot: process.cwd(), tmuxSocketName: "ccx" });

mux.create_project("my-project", process.cwd());
mux.spawn_agent("codex", "codex-1");
mux.send_to_agent("codex-1", "hello");

const { output } = mux.read_agent_output("codex-1", 50);
console.log(output);

📚 Documentation

MCP tools

The MCP server exposes these tools (names match the AgentMux methods):

Project management

• create_project({ name, cwd? })
• list_projects({})
• switch_project({ name })
• close_project({ name })

Agent management

• spawn_agent({ type, name, options? })
• list_agents({ project? })
• attach_agent({ name })
• interrupt_agent({ name })
• kill_agent({ name })

Agent I/O

• send_to_agent({ name, text })
• read_agent_output({ name, lines? })
• send_to_window({ window, text })

API (library)

Naming rules

Project names and agent names must match:

• ^[a-zA-Z0-9][a-zA-Z0-9._-]{0,127}$

new AgentMux(options?)

• options.projectRoot (string): base dir used for registry/log storage; default process.cwd()
• options.tmuxSocketName (string): tmux socket name for tmux -L; default ccx
• options.tmuxBinary (string): tmux binary; default tmux

Project management

• create_project(name, cwd?) → { project, cwd, created, storageDir }
• list_projects() → Array<{ project, cwd, sessionExists, hasRegistry, current }>
• switch_project(name) → { project, switched, clients }
• close_project(name) → { project, killed, clearedCurrent }

Agent management

• spawn_agent(type, name, options?) → agent entry (tmux window + registry + log path)
  • type: one of claude / codex / gemini
  • options.cwd (string): agent working dir (defaults to project cwd)
  • options.command / options.cmd (string): override executed command
  • options.args (string[]): extra args appended to defaults (or used with custom command)
• list_agents(project?) → { project, agents: [...entries with alive:boolean] }
• attach_agent(name) → { project, agent, attached, clients }
• interrupt_agent(name) → { project, agent, interrupted } (sends Ctrl+C)
• kill_agent(name) → { project, agent, killed }

Agent I/O

• send_to_agent(name, text) → { project, agent, sent, bytes }
• send_to_window(window, text) → { project, window, sent, bytes }
• read_agent_output(name, lines=200) → { project, agent, lines, output }

🔧 Technical Details

tmux provides durable sessions you can inspect manually, and a simple way to capture rendered terminal output—making it a practical substrate for orchestrating interactive CLI agents.

📄 License

MIT License

Copyright (c) 2026 Agent Mux Contributors

Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions:

The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software.

THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.

Note: I have a new version named Hydra, which is more powerful. See details at https://github.com/JinchengGao-Infty/Hydra?tab=readme-ov-file.