Skip to content
How It Works

See how a 30-minute retest becomes a 5-minute review

Submit a reported web finding with reproduction steps in about five minutes. The agent runs the retest unattended and hands back a Fixed or Not Fixed verdict with full HTTP evidence to review.

See the Retest Loop
riftx.io/retests/new
Submit a retest

Replay a reported finding against the live target.

https://example.com/search
Reflected XSS
primary input

Navigate to /search. Enter a script payload in the q parameter. The value reflects unescaped into the results heading and the alert box fires.

The reporter's notes are the only substantive input. Any attack strings to try live here as prose. No separate payload field.

Attach original report (optional)PDF, HAR
~5 min to submitRun retest

The submission artifact. Steps-to-reproduce, target, and workflow context enter the system in a way pentesters can read quickly and trust.

5 min

of your time per retest

HTTP traces + recording

evidence bundle attached

Needs Review

returned when unsure, instead of a guess

Reproduction steps are the only input

There is no separate payload field. The reporter's steps to reproduce are all RiftX works from, and any attack strings they want tried live inside those notes as prose. RiftX replays what was reported. It never hunts for new issues.

Why this matters: the system works from what the pentester actually did, not from a perfect proof of concept that may never have been documented in the ticket.

POST /retests
{
  "target_url": "https://example.com/search",
  "vulnerability_type": "reflected_xss",
  "description": "XSS in search parameter",
  "steps_to_reproduce": [
    "Navigate to /search",
    "Enter payload in the q parameter",
    "Observe alert box fires"
  ]
}

The interpretation layer

RiftX reads the steps, identifies the likely vulnerability profile, and chooses a retest strategy.

The goal is not to invent a new attack path. It is to replay the reported path faithfully and capture evidence cleanly.

reported finding | replay | judge | seal evidence

Agent Loop

The ReAct retest loop

Observe, interpret, act, and validate. The agent captures raw evidence at each step and keeps itself honest by checking for state change before it moves on.

01

Observe

The agent captures DOM state, console output, network activity, and browser events before it commits to the next action.

02

Interpret

The agent reads the observation for what the target actually did: dialogs, timing, response artifacts, and headers. It captures the raw evidence rather than trusting a pattern match.

03

Act

Based on the current phase, the agent chooses the next browser action: click, type, submit, wait, or check the result.

04

Validate

Each action is checked for state change. If the page stalls or diverges, the loop adapts instead of guessing.

A retest artifact in motion

isolated
  1. 01

    Observation captured

  2. 02

    Evidence interpreted

  3. 03

    Browser action selected

  4. 04

    State change validated

retest logrunning
00:00page loaded from reproduction steps
00:41goal set: reproduce reflected_xss from steps
01:36network and DOM baselines stored
03:12browser action validated for state change
04:24evidence bundle prepared
04:58verdict ready: Fixed

Evidence is part of the verdict, not an afterthought

Every retest bundle is designed to be reviewed by a human pentester after the agent finishes.

HTTP Traces

Full archive of every request and response

Screen Recording

Recording of the browser execution

HTML Report

Request/response visualization for your report

Integrity Seal

HMAC-SHA256 sealed evidence integrity

cheaper than a manual retest

5 min

of your hands on time per retest

Every

verdict independently rechecked by an always on auditor

Exactly three outcomes

No soft maybes and no vague probability language. The agent returns Fixed, returns Not Fixed, or stops honestly at a review boundary.

Fixed

The reported vulnerability could not be reproduced. The fix appears effective.

Not Fixed

Vulnerability confirmed with reproducible evidence. The original issue persists.

Needs Review

The agent hit a safety limit or could not capture clean evidence, so it returned Needs Review instead of guessing.

When RiftX cannot reach a confident verdict, it returns Needs Review. It does not guess to create a prettier report.

Built in safety limits

Every retest runs inside hard boundaries. If any limit is hit, the system terminates rather than improvising.

$10
Max Cost
per retest
25 min
Max Time
execution window
25
Max Actions
browser actions
20
Max LLM Calls
per retest

API first from submission to result

Submit retests programmatically, track state cleanly, and attach the evidence bundle back into the rest of your tooling.

curl: submit retest
curl -X POST https://api.riftx.io/jobs \
  -H "X-API-Key: vt_live_..." \
  -H "Content-Type: application/json" \
  -d '{
    "target_url": "https://example.com/search",
    "vulnerability_type": "reflected_xss",
    "steps_to_reproduce": [
      "Navigate to /search",
      "Enter payload in the q parameter",
      "Observe alert box fires"
    ]
  }'
Response
{
  "job_id": "job_7f3a2b1c",
  "status": "queued"
}

Fits the tools your team already uses

RiftX receives reported findings and returns Fixed or Not Fixed verdicts through the systems your team already relies on.

01

Vulnerability Reported

Jira / Burp Suite / API

02

RiftX Receives

Goal setting & planning

03

Browser Executes

Real Playwright retest

04

Verdict Returned

Evidence back in your tools

available

REST API

Submit retests, poll results, and retrieve evidence programmatically. JSON in, JSON out.

available

Webhooks

Get notified in real time when a verdict is ready. Push results to any endpoint.

most requested

Jira

Create retested finding tickets automatically, with HAR, MP4, and confidence score attached.

most requested

Slack

Team notifications when a retest completes. Verdict, confidence, and link to evidence.

most requested

Burp Suite

Right click a finding in Burp and send it to RiftX for an autonomous retest.

most requested

ServiceNow

Enterprise ITSM ticket creation with full evidence chain for compliance workflows.

Private Beta

Get your pentesters out of the retesting loop

Your pentesters should be finding vulnerabilities and writing reports, not manually retesting the same XSS for the third time this month.

Limited beta spots. No credit card. Set up in one session.