gemini-cli s

Gemini CLI MCP Servers extend the Gemini CLI with two built-in servers: PromptGen and SuperAgent. PromptGen helps refine user prompts by applying structured prompt engineering best practices, making it easier to generate high-quality prompts for various tasks. SuperAgent acts as a powerful orchestrator that follows an Analyze -> Decide -> Formulate -> Delegate loop to tackle complex goals by dispatching specialist agents. To use these servers, install them through the Gemini CLI using the provided fastmcp commands, then invoke them directly in the Gemini CLI with their respective slash commands (e.g., /promptgen and /superagent) to process your input prompts or high-level goals. Ensure you have a working Python environment managed by uv (the uv tool) as described in the setup sections, since the servers run as Python-based MCPs under this project.

Prerequisites:

• Python 3.8+ and uv (environment/package manager) installed
• Git installed
• Access to the Gemini CLI environment and the fastmcp tool

Installation steps:

1. Clone the repository and set up the project: git clone https://github.com/your-username/gemini-cli-mcp-servers.git cd gemini-cli-mcp-servers

2. Set up a Python virtual environment using uv and install base dependencies: uv venv source .venv/bin/activate uv pip install -e .

3. Install individual MCP servers using fastmcp: fastmcp install gemini-cli promptgen_server/main.py --name promptgen fastmcp install gemini-cli superagent_server/main.py --name superagent

4. Configure any extra requirements for SuperAgent (temporary directory and settings): mkdir -p ~/tmp/gemini-tasks/ /directory add ~/tmp/gemini-tasks/ (or update settings.json in ~/.gemini/ with the appropriate context includeDirectories path)

5. Run and test: Use the Gemini CLI commands to interact with the servers, e.g., /promptgen 'your prompt idea' or /superagent 'your high-level goal here'.

Notes and tips:

• SuperAgent requires a dedicated scratch directory. Create ~/tmp/gemini-tasks/ and register it with Gemini CLI so scratch files are written outside the current working directory.
• If you run into path or permission issues, ensure your virtual environment is activated (source .venv/bin/activate) and that uv can access pip in your environment.
• The PromptGen server provides structured prompt refinement, which can significantly improve the quality and relevance of downstream results.
• The SuperAgent server implements a loop: Analyze -> Decide -> Formulate -> Delegate to specialized agents, enabling complex, multi-step task execution.
• These servers are designed to be installed and run via the Gemini CLI ecosystem using fastmcp commands as shown in the README; replace placeholder paths with the actual deployed module paths if your project layout differs.