Can the agent prove the page reached the right state?
Back to Studio
Studio / Research
Browser Agent Evaluation Kit
Browser-agent tasks for navigation, form filling, extraction, screenshot QA, and resilient recovery from UI changes.
Live Studio demo
Browser agent evaluation kit
Evaluate whether a browser agent actually reached the target state: navigation, extraction, recovery, screenshot evidence, and handoff risk.
Extraction / Medium
Pricing page extraction
Navigate a provider pricing page, extract input/output token prices, and return source-linked structured data.
navigation success
evidence extraction
state verification
State proof
67
Recovery
50
Screenshot
74
Handoff risk
41
Browser run comparison
Browser agents extract stale snippets or fail to distinguish input, cached input, and output pricing.
Browser Operations Suite
Frontier reasoning model73
Accepted / 6150msMinor formatting cleanup needed.
Fast mid-tier model62
Accepted with review / 3520msMissed cached-input distinction.
Open-weight local model44
Partial / 5480msMixed plan names with API prices.
Small routing model28
Rejected / 1760msCould classify the page type only.
Evidence required
pricing table
source URL
structured output
Top trace proof
Returned input, cached-input, and output prices with the source URL preserved for review.
open pricing pageextract pricing tablenormalize fields
Deployment readout
Strict proof is enabled: this browser workflow should remain in supervised mode until confirmation-state checks and screenshot capture are stable.
How it works
Each Studio surface is designed as a practical operating loop: capture the buyer problem, run measured evidence, and return a decision artifact that can be acted on.
Current demo state
Research kit with browser-operation scoring; real authenticated workflow packs can be built for consulting clients.
1
Define realistic browser jobs with target URL, expected state, blocked shortcuts, and screenshot evidence.
2
Run agents through navigation, extraction, form fill, confirmation, and recovery scenarios.
3
Capture DOM state, screenshot checks, console warnings, timeout behavior, and human handoff points.
4
Publish a robustness report by site pattern rather than a single aggregate browser score.
Buyer questions
These are the questions the product needs to answer before someone deploys, buys, or scales the system.
What happens when a modal or validation error appears?
Which workflows are stable enough for automation?
Where should a human remain in the loop?
Deliverables
What a buyer gets
Browser task report
Screenshot evidence
Selector fragility map
Handoff recommendation
Connected evidence
Read the benchmark trail
Studio packet
Buyer-ready demo packet.
This generated packet gives Sanjay and Saujas a consistent follow-up artifact for demos, consulting calls, and product conversations.
Next build step
Turn this Studio surface from a populated product brief into a live demo by wiring real run data, screenshots, and client-approved examples into the same page.