home / mcp / boring mcp server

Boring MCP Server

Provides an MCP server to run Boring’s autonomous coding partner with optional Shadow Mode and RAG features.

Installation

Add the following to your MCP client configuration file.

Configuration

{
  "mcpServers": {
    "boring206-boring-gemini": {
      "command": "python",
      "args": [
        "-m",
        "boring.mcp.server"
      ],
      "env": {
        "BORING_MCP_MODE": "1",
        "PROJECT_ROOT_DEFAULT": "."
      }
    }
  }
}

Boring MCP Server lets you run Boring’s autonomous coding partner locally or via a deployment platform. You can install the MCP, configure it to run with your preferred client, and then use practical prompts to guide AI-assisted coding, verification, and project scaffolding with built-in safety (shadow mode) and memory features.

How to use

You interact with the MCP server through an MCP client. Start the server using one of the provided configurations, then connect your client to the local or remote endpoint. Once connected, you can invoke prompts to start a new project, auto-fix linting, generate semantic commits, or build a complete full-stack application. Shadow Mode protects destructive operations, so critical actions require explicit approval when enabled.

How to install

Prerequisites depend on the installation method you choose. Ensure you have the required runtime tools before proceeding.

Option 1 — Smithery (recommended) install flow. Run this in your terminal.

# Install on Smithery using the MCP client
npx -y @smithery/cli@latest install boring/boring --client gemini-cli

Option 2 — Local Python (pip) installation. Install the package with optional features.

pip install boring-aicoding
# Full installation with all features
pip install "boring-aicoding[all]"
# Specific extras
pip install "boring-aicoding[mcp]"     # MCP server + RAG
pip install "boring-aicoding[vector]"  # Pure RAG/Vector search

Option 3 — Clone from GitHub (fallback). This is best if you prefer developing locally.

git clone https://github.com/Boring206/boring-gemini.git
cd boring-gemini
pip install -e .

Configuration

Choose a configuration method that matches your installation flow. The MCP server entries below show two explicit options: one for Smithery and one for Local pip. You can adapt these snippets to your environment.

{
  "mcpServers": [
    {
      "name": "boring_smithery",
      "type": "stdio",
      "command": "npx",
      "args": ["-y", "@smithery/cli", "run", "@boring/boring", "--config", "{}"]
    },
    {
      "name": "boring_pip",
      "type": "stdio",
      "command": "python",
      "args": ["-m", "boring.mcp.server"],
      "env": {
        "BORING_MCP_MODE": "1",
        "PROJECT_ROOT_DEFAULT": "."
      }
    }
  ]
}

Shadow Mode and safety

Shadow Mode protects you from destructive AI operations. Start in a permissive state and progressively enable stricter safety as you gain confidence. The available levels are DISABLED, ENABLED, and STRICT, with STRICT requiring explicit approval for all writes.

boring_shadow_mode(action="set_level", level="STRICT")

Usage tips

- Start a new project with AI guidance using the vibe flow to scaffold a repository and setup. Use /vibe_start to begin. - Auto-fix code quality with /quick_fix to resolve linting and formatting issues. - Generate meaningful commit messages with /smart_commit to capture intent. - Build a complete full-stack app with /full_stack_dev to test the end-to-end workflow.

Security and governance

Rely on Shadow Mode for safety during high-risk actions. Maintain a log of human approvals for critical steps. Regularly update dependencies and run multi-language linting and security scans as part of your CI/CD workflow.

Available tools

boring_shadow_mode

Controls Shadow Mode levels (DISABLED, ENABLED, STRICT) to govern AI destructive actions.