CLI Agent

Unified MCP server for CLI AI agents offering Codex, Gemini, Claude, and OpenCode tooling with optional GUI and per-request isolation.

python

GitHub Stars

python

Language

4 months ago

First Indexed

3 weeks ago

Catalog Refreshed

Documentation & install

Readme and setup notes from the catalogue, plus a client-ready config you can copy for your MCP host.

Installation

Add the following to your MCP client configuration file.

Configuration

View docs

{
  "mcpServers": {
    "shiharuharu-cli-agent-mcp": {
      "command": "uvx",
      "args": [
        "cli-agent-mcp"
      ],
      "env": {
        "CAM_GUI": "true to enable GUI dashboard",
        "CAM_DEBUG": "true or false (enable debug in responses)",
        "CAM_GUI_KEEP": "keep GUI on exit",
        "CAM_LOG_DEBUG": "true to write debug logs to a temp file",
        "CAM_GUI_DETAIL": "GUI detail mode",
        "CAM_SIGINT_MODE": "cancel / exit / cancel_then_exit",
        "CAM_SIGINT_DOUBLE_TAP_WINDOW": "seconds for double-tap exit"
      }
    }
  }
}

You have a unified MCP server designed for CLI AI agents. It exposes a single interface to invoke multiple CLI tools (Codex, Gemini, Claude, and OpenCode) through a common MCP endpoint, with optional GUI monitoring and per-request execution contexts for safe concurrent use.

How to use

You integrate this server with your MCP client by configuring a local stdio MCP connection. The server supports running as a local process that your MCP client can talk to directly, and you can enable a GUI dashboard for real-time task monitoring. Use this setup to invoke different CLI agents for code analysis, UI design, feature implementation, and full-stack prototyping through a consistent interface.

How to install

Prerequisites: you need Python and the ability to install Python packages, along with the MCP client tooling you will use to connect to the server.

# Install from PyPI when available
uvx cli-agent-mcp

# Install from GitHub (recommended for development)
uvx --from git+https://github.com/shiharuharu/cli-agent-mcp.git cli-agent-mcp

# Install from source in editable mode for development
uvx --from /path/to/cli-agent-mcp cli-agent-mcp

# Or install with pip in editable mode
pip install -e .

Configure and start using the MCP client

Configure your MCP client with a stdio connection that points to the local server. The following example shows how to register the server so your client can start it and communicate over standard input/output.

{
  "mcpServers": {
    "cli_agent_mcp": {
      "type": "stdio",
      "name": "cli_agent_mcp",
      "command": "uvx",
      "args": ["cli-agent-mcp"]
    }
  }
}

Run in a typical environment

You may also run directly via a package manager workflow as shown for development or deployment. If you need to enable debug logging and verbose output for troubleshooting, you can set environment variables before starting the server.

# Enable debug info in responses
export CAM_DEBUG=true

# Enable detailed log file
export CAM_LOG_DEBUG=true

# Optional: disable GUI to run headless
export CAM_GUI=false

Available tools

codex

Invoke Codex CLI agent for deep code analysis and review

gemini

Invoke Google Gemini CLI agent for UI design and analysis

claude

Invoke Anthropic Claude CLI agent for code implementation

opencode

Invoke OpenCode CLI agent for full-stack development

Built by

VeilStrat

AI signals for GTM teams