HTTP Archive Analyzer
A HAR recording can show why a page felt slow, but it can also contain session cookies, authorization headers, full URLs and bodies. Open a local HAR 1.2 file to compare request start offsets, blocked/DNS/connect/send/wait/receive phases, known wire bytes and review candidates. The visible and downloaded result contains only request numbers, fixed categories and numeric measurements.
Key features
- Calculate a phase-accurate waterfall and distinguish elapsed wall time from the sum of overlapping request phases
- Show known response header plus body transfer bytes; preserve unknown-size counts rather than inventing bytes
- Flag HAR cache records, 304 responses and cache-control evidence as review candidates
- Group repeated successful GET requests with exactly matching URL and body size internally, exporting only anonymous request numbers
- Download a strict allowlist JSON report without source identifiers, headers, cookies or bodies
How to use
- Choose a local HAR 1.2 file of at most 8 MB and 500 entries, or load the built-in safe example.
- Run the analyzer and inspect phase totals and the request-number waterfall.
- Review candidate badges and warnings for unknown sizes or inconsistent declared timing.
- Export only the sanitized numeric report when useful; keep the source HAR private.
Use cases
- Find whether waiting or receiving dominates a recorded page load
- Compare a burst of overlapping requests against actual elapsed time
- Spot repeated transfers or cache metadata worth investigating without exposing URLs
- Share timing evidence without distributing credentials or response bodies
Frequently asked questions
Does this send the HAR to a server?
No. The selected file is parsed in browser memory; no endpoint receives its contents. Only a numeric and fixed-label report is rendered or exported.
Does a duplicate badge prove unnecessary traffic?
No. It identifies repeated GET 200 entries with exactly matching URL and response body size within this HAR. Different request headers or content can still explain the requests. Investigate in your private source file.
What counts as transferred bytes?
Only the sum of HAR response headersSize and bodySize when both are known. A -1 size is unknown, not zero. Decoded content.size is not used as network transfer.
Why does the phase sum differ from wall time?
Requests can overlap, so summed request phases can exceed elapsed wall time. The ssl field is included inside connect and is displayed as information only, never added again.
Is every browser HAR supported?
This bounded profile accepts HAR version 1.2 JSON with valid HTTP(S) request entries, supported timing fields and up to 500 entries. Browser extensions and malformed recordings may be rejected; it is not a network audit or complete HAR conformance validator.
Privacy
Analysis stays in this browser. The report never includes request URLs, hostnames, paths, query strings, headers, cookies, body text, creator metadata, page titles or source filename. Review your own original HAR separately before sharing it.
Comments & questions