How Monitoring Works

Varen uses a two-tier detection system to catch changes across any API documentation format.

Detection tiers

Tier 1: Deterministic diffing (~40% of APIs)

For APIs that publish an OpenAPI/Swagger specification, Varen uses oasdiff — an open-source, deterministic diff engine. This catches:

• Added, removed, or modified endpoints
• Parameter changes (name, type, required status)
• Response schema modifications
• Authentication changes
• Deprecation annotations

No AI is involved. The diff is exact and reproducible.

Tier 2: AI-powered analysis (~60% of APIs)

For APIs with only HTML or markdown documentation (no structured spec), Varen uses Claude to:

1. Extract structured data from the raw documentation
2. Compare snapshots to identify meaningful changes
3. Classify changes by severity
4. Filter noise — whitespace changes, reordering, and cosmetic updates are ignored

Polling cadence

Catalog APIs are shared across all users — one scan serves everyone who monitors that API.

Snapshot lifecycle

Each scan creates a snapshot of the API's current documentation state:

1. Varen fetches the documentation URL
2. The content is parsed and normalized
3. It's compared against the previous snapshot
4. Any differences are recorded as change events
5. Changes are classified by severity and stored

You can view the before/after diff for any change in the dashboard.

Rate limiting and politeness

Varen respects robots.txt and rate limits when scraping. Documentation is fetched once per scan cycle — we don't hammer endpoints.