DefensiveAI

Tool · runs on your machine

Run the analysis yourself.

The same recursive decomposition and roll-up synthesis we use in every engagement, packaged to run on your machine against your own repository. Your code never leaves your computer. You bring the model API key.

Download

Public builds are coming. Request early access

macOS · Linux · Windows · No support commitment.

What it does

Reads everything, keeps everything, hands you the tree.

01

Reads every line

Splits the codebase into units small enough to read in full, delegates each to a copy of the analysis agent, and synthesizes upward.

02

Writes everything to disk

The full observation tree, as YAML. Nothing is discarded; every leaf report is there to search.

03

Generates a report you open in a browser

A static page: the roll-up at the top, drill down to every leaf. Yours to keep, share, or ignore.

What it doesn’t do

It doesn’t validate. The output is every observation the analysis recorded — things that look wrong, fragile, or merely odd. Most of them will be nothing. Some of them will matter. Telling the difference, and seeing how they combine, is the validation and relevance work we do for clients.

What an engagement adds

Requirements

  • A repository on disk.
  • An API key for a supported model provider.
  • Time and tokens proportional to codebase size.

This is the checklist an attacker’s agent runs, pointed at your own code first.

Start with a free evaluation.

A top-level pass over your codebase and a 30-minute conversation about what we saw. No charge, no obligation.