Open-source project / Local AI tooling
OpenSecurityMCP.
Security workflows, with local inference.
An Ollama-powered adaptation of SecurityMCP with a CLI and authenticated HTTP API.
THE IDEA
Why I built it.
OpenSecurityMCP brings the document-security demo to a locally hosted model, with explicit controls for document edits and an API for testing integrations.
IN PRACTICE
What it does.
Run tool-capable local models through Ollama.
Access the demo through an interactive CLI or bearer-authenticated API.
Opt into edits explicitly, with enforcement in the tool router.
Validate inputs and bound tool loops and request duration.
CONTEXT & SCOPE
Where it fits.
This is an in-memory document-security demonstration, not a disk scanner or compliance product. With a local model and the default local host, inference stays on the machine.