Introducing deepsec: The security harness for finding vulnerabilities in your codebase
Not a pilot. Vercel just open-sourced an AI security agent that scales to 1,000+ concurrent scans—and it catches what static analysis misses.
Why it matters
Vercel shipped a production-ready coding agent product (deepsec) that automates vulnerability detection across large codebases using Claude Opus and GPT-5.5. This is a concrete application of AI agents solving enterprise security workflows at scale, with real deployment data from Vercel's own monorepos.
The key facts
8 to knowdeepsec open-sourced as security harness powered by coding agents
Runs on local infrastructure; supports Claude Opus 4.7 and GPT-5.5 for inference
Scales to 1,000+ concurrent sandboxes via Vercel Sandboxes for parallel execution
10-20% false positive rate; includes revalidation step to reduce noise
Tested on Dub.co (open source) and Vercel's own monorepos; identified subtle auth edge cases
Ships with plugin system for custom scanners (regex matchers, auth model tuning)
Works with both standard and 'cyber' versions of Claude/OpenAI models
Multi-step workflow: Scan → Investigate → Revalidate → Enrich → Export
Go to the source
Vercel Blogvercel.com
Publisher excerpt: Today we’re open sourcing : a security harness powered by coding agents. It runs on your own infrastructure and surfaces hard-to-find issues in large codebases. deepsec You can run on your laptop without setting up a cloud service for privileged source code access. For inference, you can use your…