argus

Argus MCP provides AI-powered code review with a zero-trust approach. It orchestrates multiple AI models (such as GLM 4.7, Gemini 3 Flash Preview, and MiniMax M2.1) to inspect code and produce structured feedback categorized as Must Fix, Should Fix, and Suggestions. With language-aware checks, security reviews (OWASP), and architecture evaluations, Argus supports single-file, git-diff, and multi-file review modes. To use it, run the MCP server via your environment and connect from any MCP client (Windsurf, Claude Desktop, Cursor, or other MCP-compatible tools). The tool automatically determines the appropriate review scope based on your IDE context and the files under consideration, returning a detailed, actionable report.

Key capabilities include:

• Zero-Trust code review that prioritizes security and correctness
• Mode selection (Single File, Git Diff, Multiple Files) based on context
• Language-specific checks for Python, JavaScript, TypeScript, and more (10+ languages)
• Smart retry and model fallback to ensure robust results
• Caching of results to speed up repeated checks
• Clear output formatting with exact file references and remediation steps

Prerequisites:

• Python 3.11+ installed
• A Python virtual environment workflow (venv) or similar
• Git available to clone the repository
• API keys for GLM and OpenRouter (as required by configuration)

Step-by-step installation:

1. Clone the repository and set up a virtual environment

git clone https://github.com/lokafinnsw/argus-mcp.git
cd argus-mcp
python3 -m venv venv
source venv/bin/activate  # On Windows: venv\Scripts\activate

2. Install dependencies

pip install -r requirements.txt

3. Create and configure environment variables

cp .env.example .env

Edit .env to include your API keys (GLM_API_KEY, OPENROUTER_API_KEY) and default model (DEFAULT_MODEL).

4. Run the Argus MCP server

python server_v2.py

5. Connect via an MCP client (Windsurf, Claude Desktop, Cursor) using the configured mcp_config.json path, typically pointing to "/absolute/path/to/argus-mcp/venv/bin/python" with the server script as an argument.

Tips and notes:

• Ensure the absolute installation path is correctly referenced in your MCP client configurations (replace /absolute/path/to/argus-mcp).
• Store API keys securely in the .env file and add it to your .gitignore.
• If you encounter model timeouts or rate limits, leverage Argus' multi-model fallback and caching to reduce latency.
• When updating dependencies or models, re-run installation steps and verify client reloads (Windsurf, Claude Desktop, Cursor).
• For troubleshooting, check server_v2.py logs and environment variable loading errors; verify that the Python virtual environment is active when launching.