Opsec runs a fleet of parallel AI security agents that find exploits and auto-open fix PRs

AccBalanced · x · 2026-11-04

Opsec launched a security AI product for solo founders and small teams, pitched as "hack yourself before they do."

Splits a repo into ten attack surfaces—from gateway token issuance to plugin, MCP and skill supply chains—and assigns each one its own agent, all working in parallel
Each agent first learns how its part of the app works, then reads code, runs searches and commands, and follows its own leads; you can open any agent mid-run to see which files it is reading and what it is executing
Hunts run in the background, monitor every commit, deploy and config change, and every finding is validated with a working exploit; fixes ship as pull requests
The demo shows a confirmed reentrancy bug in Vault.sol and an IDOR in GET /api/orders/:id, each with an auto-opened fix PR

Positioned as big-company-grade security coverage across code, contracts, web apps and cloud infrastructure—no security team required.

More from coding & agent