YC X26 Applicant

Autonomous bug detection
for mobile apps

Give it a goal in plain English. It explores your app, finds bugs, and generates reproduction steps — no test scripts needed.

swarm explore
$ swarm explore --goal "Verify search quota is enforced after 5 queries" --app com.myapp
Starting exploration...
Step 1: Launching app...
Step 2: Navigating to Search tab...
Step 3: Typing "test query"...
Step 4: Tapping send button...
...
Step 12: Executing 6th search query...
RESULT: BUGS FOUND
[CRITICAL] App crash on 6th search query (SIGABRT)
[WARNING] Send button unresponsive for 4.2s
Reproduction Steps:
1. Launch app
2. Tap "Search" tab
3. Type "test query" → tap send (repeat 6x)
4. App crashes on 6th query
Evidence: screenshot_012.png, crash_log.txt

See it in action

Watch Swarm autonomously explore an app and find a bug in under 60 seconds.

Demo video (60s)

Terminal + iOS Simulator side-by-side. Real bug found in a real app.

Mobile QA is broken

Every mobile team faces the same impossible tradeoffs.

Manual QA is slow

A single human tester covers ~20-30 flows per day. Most teams can't afford to test every feature after every deploy.

Test scripts are brittle

Conventional automation (Detox, Appium, XCUITest) breaks every time the UI changes. Teams spend more time fixing tests than writing features.

Code agents ship bugs

AI-generated code is proliferating. Nobody is verifying whether the features actually work as intended on the device.

How it works

Three steps. Zero test scripts.

1

Describe what to test

swarm explore \
  --goal "Verify the quota is enforced after 5 search queries" \
  --app com.myapp

No test scripts. No selectors. Just describe what you want verified.

2

AI explores autonomously

Exploring...
tapOn "Search" tab
inputText "test query"
tapOn send button
Running oracles...

Navigates like a real user. Handles auth dialogs, adapts when things fail.

3

Get a bug report

BUGS FOUND: 2
[CRITICAL] Crash on 6th query
[WARNING] Button lag 4.2s
+ screenshots
+ crash logs
+ repro steps

Every bug comes with exact steps to reproduce and evidence.

Built different

Not another test automation tool. An autonomous verification layer.

Zero test scripts

Describe goals in English, not fragile selectors. Survives UI changes.

4 bug detection engines

Crashes, visual defects, performance regressions, behavioral anomalies — detected automatically.

0.5s per action

Persistent Detox server, not 10-15s per CLI spawn. 20x faster exploration.

Minimal repro steps

Every bug comes with exact steps to reproduce. Hand it to any engineer.

Learns your app

Caches screen maps across runs — gets faster and smarter over time.

CI-ready JSON output

Designed for code agents and automated pipelines, not just humans.

The Future of QA

Built for the code agent era

Code agents write features. Swarm verifies they work.

AI-generated code is shipping to production faster than any team can review it. Swarm sits at the end of the pipeline — after the code agent commits, Swarm explores the app on a real device and reports what actually broke. Feature-level truth, not line-level linting.

Code Agent
CI Build
Swarm
Bug Report
Auto-fix

Under the hood

Technical details for the curious.

Vision-first exploration

Gemini vision analyzes screenshots at each step. Decides actions, tracks verification plans, adapts in real-time.

4 specialized oracles

CrashOracle (15+ crash signatures), VisualOracle (UI defects), PerformanceOracle (latency), BehaviorOracle (loops, stuck screens).

Persistent Detox server

Node.js HTTP server keeps a single Detox session alive. 20-30x faster than spawning per action.

App model caching

Learns screen structure across runs. Provides context to the LLM for faster, smarter navigation.

Radically affordable

Run 1,000 explorations for the cost of one QA engineer's lunch.

Per exploration run
$0.01-0.02
~40-50K tokens at Gemini Flash rates
Compare to traditional QA:
$150-200K
QA engineer / year
20-30
flows / day

Where we're going

From CLI to cloud. From React Native to everywhere.

Shipped

Single-app CLI

iOS React Native apps, Gemini-powered exploration, local execution

Now

CI Integration

GitHub Actions, batch test generation, parallel execution

Q3 2026

Cross-platform

Android support, web app support, multi-LLM (Claude, GPT-4o)

Q4 2026

SaaS Platform

Upload APK/IPA, get bug reports. No local setup required.

2027

Continuous Monitoring

Run on every PR, block merges that break features. The new CI/CD.

Trusted by mobile teams

Real bugs caught. Real time saved.

127
bugs caught before production
8
teams in private beta
2.4k
exploration runs completed
94%
bug repro accuracy

"Swarm found a payment flow crash on the first run that we'd missed for 3 weeks. Saved us from a really bad App Store review."

MK
Marcus K.
Senior Engineer, Fintech Startup

"We were spending 2 days per sprint on regression testing. Swarm covers our critical flows in 20 minutes. Game changer."

SL
Sarah L.
QA Lead, Series A Health Tech

Stop writing test scripts.
Start finding bugs.

Be the first to know when Swarm launches.

No spam. Just launch updates.