OperonSec
Request access

Vulnerability Research Agent

Finds the bugs scanners can't reach.

Autonomous agents test your app with real accounts and real browsers. Only findings that reproduce reach you.

Request accessSee how it hunts
operonsec · engagement · lab-crapiLive
Attack graph: recon and the JavaScript map lead to accounts A and B, whose differential test produces a verified finding.targetreconjs mapaccount Aaccount BcoverageA/B diffverified
  1. orchestratorscope loaded · lab-crapi
  2. recon214 routes, 3 roles mapped
  3. operator Asigned up, session captured
  4. differB read A's order · candidate
  5. validatorcold start · reproduced
  • Scope enforced in code
  • Replayed before reported
  • You submit, not the agent

01 · How it works

Target in. Proof out.

  1. 01

    Scope

    Enforced before any tool runs

  2. 02

    Map

    Hosts, routes and roles

  3. 03

    Test

    Two accounts, real browsers

  4. 04

    Prove

    Cold validator, exact replay

  5. 05

    Report

    You review. You submit.

02 · The squads

One orchestrator. Narrow squads.

Each squad has its own instructions, tools and container image.

  • Orchestrator
  • Recon
  • Exploit
  • Verify
  • Report
  • Analysis (coming next)
  • Coverage (coming next)
  • Chain (coming next)

Active now Coming next

03 · Proof

Proof, not guesses.

Every candidate is tested across two accounts, then reproduced by a validator that never saw the hunt.

Two accounts, one truth

  1. APOST/api/v2/orders201 · id 4471
  2. BGET/api/v2/orders/4471200 · owner A
  3. candidateB read A's order

The finder never approves

  • Recon
  • Exploit
Validator · cold

reproduced 3/3

No transcript, no reasoning. Just the claim and the target.

Sample · OWASP crAPI labVerified

Broken object-level authorization

Medium · CVSS 6.5CWE-639replayed 3/3

GET /api/v2/orders/{id} · account B reads account A's order

validator: reproduced from empty context · human: pending

04 · Architecture

Root inside the box. Nobody outside it.

Agents get root inside three nested rings. A compromised agent stays trapped.

Cloudflare and Caddy. TLS end to end; the origin answers only through Cloudflare.

The orchestrator runs in a non-root outer container (Sysbox). No host Docker socket, no privileged mode.

One short-lived container per task. Root inside maps to nobody on the host.

Its own network per engagement. No metadata service, no internal hosts; only in-scope traffic leaves.

05 · Controls

Bounded by code, not good intentions.

  • Scope before autonomy

    Out of scope is blocked, not discouraged.

  • Target content is data

    Nothing a target says can change the rules.

  • Evidence or it didn't happen

    Reports cite stored requests and screenshots.

  • You make the last call

    Nothing is submitted without you.

06 · Early access

Request early access.

Opening gradually to researchers who test authorized targets.

Early access opens soon. Not collecting submissions yet.

07 · FAQ

Questions

Is OperonSec available?

Not yet. It's an early research system, and access opens gradually.

Which targets can it test?

Only targets you are authorized to test, within their published scope and rules. Scope is enforced in code.

Does it submit reports for me?

No. It prepares a reproducible report; you review it and submit it yourself.

Which AI models does it use?

Several providers, including DeepSeek and GLM, with a model chosen per agent role.

Can I test OperonSec itself?

No. Testing of operonsec.com or any OperonSec system is not authorized.