prettyRECON Zero
prettyRECONZero
Sign inJoin waitlist
Authorized targets only

AI recon & bug hunting, from one control plane.

prettyRECON studies an authorized target, deduces which bugs are plausible from real evidence, then spawns one dedicated agent per suspected bug — and verifies every finding with a different agent.

Join the waitlistSign in

Access opens in batches. Create your account to join the list — we'll email you the moment it's your turn.

  • one agent per bug
  • independent verification
  • deterministic reports
app.shopdemo.io — prettyRECON Zero
app.shopdemo.ioRunning · Hunt · 66%
Stop
  1. Recon
  2. Strategy
  3. Hunt
  4. 4Verify
  5. 5Report
OverviewHypotheses12Findings3

Cross-account order read

CRITICALConfirmed

Sequential order IDs let any authenticated user fetch another customer's order — address, line items, and totals — with no ownership check.

14:22GET /api/orders/1042· hunter · idor

Reflected XSS in search query

HIGHCandidate

The q parameter is echoed unescaped into the results heading; a crafted value executes script in the victim's session.

14:24GET /search?q=· hunter · xss

Open redirect on login return_to

MEDIUMConfirmed

return_to accepts absolute off-domain URLs and redirects after login — usable for phishing and token theft.

14:25GET /login?return_to=· hunter · open-redirect

Role escalation via mass assignment

HIGHRefuted

Sending role=admin to the profile endpoint was rejected server-side — the field isn't bound. Verifier could not reproduce.

14:26PATCH /api/profile· hunter · mass-assignment
37+bug classes hunted
50+methodology playbooks
5pipeline phases
1agent per suspected bug
The pipeline

Five phases. One agent per suspected bug.

prettyRECON runs a deterministic pipeline. Each phase hands structured evidence to the next through a shared workspace — never through chat — so no agent's context is ever flooded.

  1. Recon
  2. 2Strategy
  3. 3Hunt
  4. 4Verify
  5. 5Report
Phase 1 / 51 agent

Recon — Maps the attack surface

Crawls with katana, pulls archives with waymore, dumps and greps the app's JavaScript, captures live API traffic, and fingerprints the stack — then writes a classified surface map.

Why it's different

Built to prove bugs, not to guess them.

Generic AI scanners hallucinate and bury you in false positives. prettyRECON is engineered around evidence, isolation, and independent proof.

One agent per bug

Every suspected bug gets its own hunter, its own authenticated browser, and a single methodology skill — isolated context, run in parallel.

Independent verification

Finding identity is orchestrator-assigned, so a bug can only be confirmed by a genuinely different agent. No self-confirmation, ever.

Reports that can't hallucinate

The report is built in code from proven findings. If it wasn't reproduced, it doesn't ship — so you never chase a bug that isn't there.

Logs itself in

The pipeline finds the login page on its own, detects the form, and seeds the session into every browser. Credentials never enter an agent's context.

Two accounts unlock IDOR

Give it a second identity and hunters switch principals to prove cross-account reads and writes — the access-control bugs scanners miss.

Scope-locked by design

Every request passes one scope choke point. Out-of-scope URLs are stored but never probed, and hunters prove bugs with the minimum PoC.

Coverage

37+ bug classes, each with a dedicated hunter.

The strategist selects from 50+ methodology playbooks — vendored from real bug-bounty tradecraft — and hands each hunter the single skill for its bug, capped so its context stays sharp.

idorxssdom-xsssqlinosqlissrfsstircelfixxecsrfcorsopen-redirectauth-bypassjwtoauthsamlsessionmfa-bypassatobusiness-logicmass-assignmentrace-conditionfile-uploadprototype-pollutiondeserializationgraphqlapi-misconfighost-headercrlfcache-poisonweb-cache-deceptionwebsocketsource-leakcloud-misconfigbrute-forcellm-ai
Framework-aware playbooks
Next.jsNode.jsLaravelSpring BootASP.NETGraphQL
See it work

The desktop app, end to end

Create a project, add an authorized target, and watch every phase, agent, and finding stream live.

The hunt

One agent per suspected bug

Every hypothesis the strategist files gets its own hunter — a dedicated agent with its own authenticated browser and a single methodology skill. They run in parallel and share nothing but the workspace.

  • Isolated context per hunter — no prompt soup
  • Its own browser, seeded with your session
  • Fires real PoCs, records the evidence
app.shopdemo.io
Hypotheses12
idor Hunted

GET /api/orders/{id}

Sequential integer IDs and no visible ownership check in the response.

agent · hunter-2

xss Claimed

GET /search?q=

User input reflected into the results heading without encoding.

agent · hunter-4

ssrf Open

POST /api/import?url=

The importer fetches a user-supplied URL server-side.

jwt Open

Authorization: Bearer …

alg header is echoed and the token isn't re-validated on refresh.

open-redirect Hunted

GET /login?return_to=

Redirect target is taken from an unvalidated query parameter.

agent · hunter-5

The proof

Every finding, independently verified

A different agent reproduces each candidate from a clean session before it can be marked confirmed. Severity, reproduction, and the verdict are all on the card — and refuted findings are kept, not hidden.

  • No agent confirms its own finding
  • Refuted candidates are dimmed, not deleted
  • The report is rendered from confirmed rows only
app.shopdemo.io
Findings3

Cross-account order read

CRITICALConfirmed

Sequential order IDs let any authenticated user fetch another customer's order — address, line items, and totals — with no ownership check.

14:22GET /api/orders/1042· hunter · idor

Reflected XSS in search query

HIGHCandidate

The q parameter is echoed unescaped into the results heading; a crafted value executes script in the victim's session.

14:24GET /search?q=· hunter · xss

Open redirect on login return_to

MEDIUMConfirmed

return_to accepts absolute off-domain URLs and redirects after login — usable for phishing and token theft.

14:25GET /login?return_to=· hunter · open-redirect

Role escalation via mass assignment

HIGHRefuted

Sending role=admin to the profile endpoint was rejected server-side — the field isn't bound. Verifier could not reproduce.

14:26PATCH /api/profile· hunter · mass-assignment
The recon

Maps the whole attack surface first

Before a single payload flies, recon crawls the app, mines its JavaScript and archives, captures live API traffic, and fingerprints the stack — then classifies every endpoint into a surface map the strategist reasons over.

  • katana crawl + waymore archives + live traffic
  • Tech fingerprint drives skill selection
  • Scope-classified: in-scope vs. stored-only
app.shopdemo.io
Overview
Target info
Target
app.shopdemo.io
Scope
shopdemo.io + subdomains
Phase
Hunt
Endpoints
214 discovered · 137 in scope
Technologies
Next.jsReactNode.jsPostgreSQLCloudflareStripe
Effort dial

Dial the depth. You control the spend.

Set the effort on every scan. Higher effort makes each agent work harder and dig deeper — it spends your monthly allowance faster, it doesn't add any. You watch what's left as a percentage of your plan.

Effort for this scan

Balanced. The default.

Heads up: Higher effort spends your monthly allowance faster — it does not add any.

Safety

Authorized targets only — and it means it.

prettyRECON sends real, working PoC payloads. Every guardrail below is structural, not a setting you can forget to turn on.

Authorized targets only

The hunt sends real PoC payloads. Point it only at systems you're explicitly allowed to test.

Minimum-PoC guardrail

Hunters prove a bug with the smallest possible action — no mass deletes, sends, or purchases, and no writing to your disk.

Credentials stay out of context

Logins run over CDP from the host. Test-account secrets never enter an agent's context or the transcript.

Locked to your scope

One scope choke point confines every request to the target's registrable domain and its subdomains.

Pricing

Priced by depth, not by seat.

Pro and Max are self-serve. Higher tiers buy deeper hunts, more hypotheses, and more parallel hunters — not just a bigger allowance.

Most popular

Pro

$50/mo

For solo hunters running engagements week to week.

Join the waitlist
Usage / month
Generous
Hypotheses / hunt
12
Parallel hunters
10
Hunt · verify turns
50 · 30

Max

$200/mo

Deepest hunts and the highest parallelism, for full-time work.

Join the waitlist
Usage / month
5× Pro
Hypotheses / hunt
25
Parallel hunters
15
Hunt · verify turns
80 · 50

Enterprise

Custom

Custom limits, seats and support. Talk to us.

Contact sales
Usage / month
Custom
Hypotheses / hunt
40
Parallel hunters
20
Hunt · verify turns
120 · 80

One-time terms of 1, 3, 6, or 12 months — not a subscription. Prices in USD. Every plan includes a monthly usage allowance that resets each period.

FAQ

Questions, answered

Early access

Join the waitlist.

prettyRECON opens in batches so every hunt gets the browsers and depth it needs. Create your account to take a place in line — no download, nothing to install today.

Join the waitlistAlready joined? Sign in
  1. 1.Join

    Register with your email. That account is your place on the list, and the same one you'll hunt with.

  2. 2.Get access

    We open accounts in batches and email you when it's your turn, along with the app for your machine.

  3. 3.Hunt

    Pick a plan, add an authorized target, and let prettyRECON hunt it end to end — then verify every finding.

Authorized targets only.