Vectra AI MCP Server exposes the Vectra integration capabilities to MCP-enabled AI tools. This server is implemented in Python and uses uv (uvx) for environment- and transport-agnostic operation, supporting multiple transport modes such as stdio (default, for Claude Desktop and similar native clients), SSE over HTTP (for web-based MCP clients), and streamable-http (recommended for production HTTP deployments). After starting the server, you can interact with Vectra AI Platform through natural language prompts to investigate threats, perform containment or remediation actions, and generate security analysis visualizations and reports. To connect an MCP client, configure the client to point to the server’s transport endpoint (for example a local stdio pipe, an HTTP SSE endpoint, or a streaming HTTP endpoint) and ensure the MCP server process is running with the correct project directory and transport settings. The README also shows how Claude Desktop can be configured to load the MCP server by adding an entry under mcpServers with the uv invocation so Claude can spawn the server as needed.

Prerequisites

• Python installed (as required by the project). Check the .python-version file for the exact version.
• uv (Python package manager) installed to run the MCP server.
• Git to clone/download the project.

Installation steps

1. Clone or download the project to your local machine.
2. Navigate into the project directory: cd /path/to/your/project
3. Create and activate a virtual environment (optional but recommended): uv venv

   Activate it (macOS/Linux)

   source .venv/bin/activate

   Activate it (Windows)

   .venv\Scripts\activate
4. Install dependencies via uv (exact versions from uv.lock are honored by uv sync): uv sync
5. Configure environment variables (example shown; adjust to your Vectra credentials): cp .env.example .env

   Edit .env to set:

   VECTRA_BASE_URL, VECTRA_CLIENT_ID, VECTRA_CLIENT_SECRET, etc.

6. Run the server: python server.py

   Transport options (examples):

   python server.py --transport stdio

   python server.py --transport sse --host 0.0.0.0 --port 8000

   python server.py --transport streamable-http --host 0.0.0.0 --port 8000

7. (Optional) For Claude Desktop integration, configure Claude to load the MCP server by adding the mcpServers entry as shown in the README, pointing to your uv executable and project directory.

Docker (optional) deployment mirrors the same environment variables and transport choices and is covered in the README under the Docker Deployment section.

Notes and tips:

• The server supports multiple transports: stdio (default), sse (HTTP SSE), and streamable-http (production-friendly HTTP).
• Ensure VECTRA_BASE_URL, VECTRA_CLIENT_ID, and VECTRA_CLIENT_SECRET are correctly set in the environment (.env) file for authentication with the Vectra AI Platform.
• If using Claude Desktop, you will typically add an MCP entry that points to your uv binary and the project directory, then run the server through Claude's UI.
• When running via Docker, you can use the pre-built images and customize transport via environment variables (VECTRA_MCP_TRANSPORT, VECTRA_MCP_HOST, VECTRA_MCP_PORT).
• If you encounter port binding issues, verify that the host and port match what your client expects and that the port is not blocked by a firewall.
• Debug mode can be enabled with VECTRA_MCP_DEBUG=true to aid troubleshooting.