Acquire
Fetch public pages through lightweight transports, managed browsers, or source-specific connectors.
Platform
One platform for reliable web data
Fetch, browser rendering, extraction, crawling, and source-specific collection share one production boundary and one explainable result model.
System design
A data path with explicit ownership.
Each layer has one job. Global quality and fallback decisions stay out of fetch and browser executors, so behavior remains inspectable as the system grows.
Route
Choose the next engine through quality, budget, identity, and capacity decisions owned by one router.
Extract
Normalize content and run typed structured extraction over fetched, rendered, or supplied HTML.
Explain
Return public-safe audit, diagnostics, artifacts, and usage alongside accepted page output.
Consistent output
Build downstream logic once.
Page results use a normalized contract whether the winning path was lightweight fetch or browser rendering.
result
├── content
│ ├── markdown
│ ├── html
│ ├── metadata
│ └── links
├── artifacts
├── audit
├── usage
└── diagnosticsLight first
Avoid browser cost when direct retrieval already meets the quality bar.
Browser when needed
Use managed browser capacity for rendering, interaction, and dynamic targets.
Bounded work
Keep time, page, retry, and cost decisions explicit in each workflow.
Public-safe evidence
Expose useful diagnostics without leaking credentials, profiles, proxies, or raw internals.
Durable jobs
Track asynchronous crawls and source collection through stable job state.
Downstream ready
Return clean content and typed data for search, analysis, and agent context.
Start building
Start with the path your workload needs.
Use one API boundary for fetch, browser, extraction, crawl, and run inspection.