qbaf-mcp-server

nckclrk/qbaf-mcp-server

3.2

If you are the rightful owner of qbaf-mcp-server and would like to certify it and/or have it hosted online, please leave a comment on the right or send an email to henry@mcphub.com.

QBAF MCP Server is a minimal, opinionated Model Context Protocol server designed to control agent access to a QBAF backend.

QBAF MCP Server

Minimal, opinionated Model Context Protocol (MCP) server that gates agent access to a QBAF backend.

Development

Create a virtual environment and install dependencies:

python3 -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt
pip install -r requirements-dev.txt

Add the source tree to your Python path when running tools locally:

export PYTHONPATH=src

Run tests:

python -m pytest

Configuration

Runtime settings are loaded from a .env file automatically if it exists in the project root. Copy .env.example to .env and fill in at least the required values before launching the server:

GRAPH_ID=your-graph-id
BACKEND_BASE_URL=http://127.0.0.1:8000
# Optional overrides
BACKEND_API_KEY=
TIMEOUT_SECONDS=10
LOG_LEVEL=INFO

To override a value temporarily, set the environment variable in your shell; explicit environment variables still take precedence over the .env file.

Local Backend (Optional)

For manual testing without a live QBAF deployment, run the built-in Starlette backend:

source .venv/bin/activate
uvicorn qbaf_mcp_server.dev_backend:app --reload

Point BACKEND_BASE_URL at the running service (default http://127.0.0.1:8000). Once the backend is up you can launch the MCP server through Inspector:

npx @modelcontextprotocol/inspector python --directory . run python -m qbaf_mcp_server.server