Nova

NOVA MCP is a MCP that leverages NOVA for prompt security

python

12

GitHub Stars

python

Language

6 months ago

First Indexed

2 months 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": {
    "fr0gger-nova_mcp": {
      "command": "python",
      "args": [
        "nova_mcp_server.py"
      ],
      "env": {
        "OPENAI_API_KEY": "YOUR_OPENAI_API_KEY",
        "ANTHROPIC_API_KEY": "YOUR_ANTHROPIC_API_KEY",
        "AZURE_OPENAI_API_KEY": "YOUR_AZURE_OPENAI_API_KEY",
        "AZURE_OPENAI_ENDPOINT": "https://<your-endpoint>"
      }
    }
  }
}

The NOVA MCP Security Gateway is a prompt-validation layer that runs first in your MCP chain. It loads Nova rules and checks each incoming prompt, allowing only authorized prompts to pass through while blocking high‑risk or jailbreak patterns. You can run it locally and connect a client to validate prompts before further processing.

How to use

Use the MCP server with any MCP client by starting the local gateway and then sending prompts through the client. The gateway evaluates prompts against configured Nova rules and responds with AUTHORIZED or NOT AUTHORIZED. Your client should always route prompts through this gateway first to ensure compliance with the defined security rules.

Key usage patterns you can implement: 1) Validate every user prompt before proceeding with higher‑risk actions. 2) Log all evaluated prompts in structured JSON for auditing and debugging. 3) Update and reload rules from the rules directory to adapt to evolving security requirements.

How to install

Prerequisites: ensure Python is available on your system. You will also need network or STDIO access depending on how you run the gateway and client.

Step 1: Move into the MCP workspace and install dependencies.

Step 2: Create an environment file with your API keys for the backends you plan to use (OPENAI_API_KEY, ANTHROPIC_API_KEY, AZURE_OPENAI_API_KEY, etc.). Place this in the nova_mcp directory.

Running the server and client

Run the MCP server from its directory to start validating prompts.

cd nova/nova_mcp
python nova_mcp_server.py

Use the reference client to interact with the MCP gateway. It will spawn the server, send prompts, and print the gateway’s response.

python client.py nova_mcp_server.py

Environment and configuration notes

Configure the rules and logging to suit your environment.

Rules directory: nova_rules/ — place your .nov rule files here.
Logs directory: logs/ — all events are logged in logs/nova_matches.log.
Environment: populate .env or export environment variables for your chosen LLM backends (examples shown in the setup). The gateway reads these variables to connect to your models.

Security and behavior notes

High‑severity or jailbreak patterns trigger NOT AUTHORIZED with rule details.
If a prompt does not match any rule, it is AUTHORIZED.
The system promotes strict prompt validation as the first step in the MCP chain to minimize downstream risk.

Managing rules and logs

Add or edit .nov files in the nova_rules/ directory. Restart the server to load changes.

JSON logs capture both authorized and blocked prompts, enabling you to audit responses and refine rules over time.

Available tools

validate_prompt

Evaluates a prompt against configured Nova rules and returns AUTHORIZED or NOT AUTHORIZED with rule details.

load_rules

Loads rule files from nova_rules/ and applies them to incoming prompts for validation.