Offline HAR Network Waterfall Analyzer

Summarizes HAR request totals, body bytes, slowest entries and HTTP errors; no timing-phase waterfall.

Har Waterfall Analyzer · Client runtimeSandboxed

Har Waterfall Analyzer Input

Har Waterfall Analyzer Output

Formatted results will render here in real time.

Har Waterfall Analyzer: JavaScript is required to run this tool in the browser.

About this offline har network waterfall analyzer

Summarizes HAR request totals, body bytes, slowest entries and HTTP errors; no timing-phase waterfall Performs protocol inspection, IP calculation, and network analysis client-side without sending probes or exposing network topology.

Protocol mechanics & network logic

Parses HAR JSON locally and shows request timing, start offsets, available blocked/DNS/connect/send/wait/receive/SSL phases, a waterfall timeline, domain/MIME/status filters, slow and large request subsets, and a redacted response-body view. Local files use incremental UTF-8 streaming and per-entry normalization; pasted text still uses JSON.parse. Normalized request details remain in memory, and peak memory/performance at the 100 MB (100,000,000-byte) limit is unverified.

Worked network diagnostic scenario

Scenario: Review aggregate results and the phase timeline for two HAR entries with timestamps and timings.

Sample input:

{"log":{"entries":[{"startedDateTime":"2026-09-01T12:00:00.000Z","time":120,"request":{"url":"https://example.com/api","method":"GET"},"response":{"status":200,"bodySize":512,"content":{"mimeType":"application/json","size":512}},"timings":{"blocked":1,"dns":10,"connect":20,"send":5,"wait":70,"receive":14,"ssl":5}},{"startedDateTime":"2026-09-01T12:00:00.250Z","time":40,"request":{"url":"https://czoa.com/assets.css","method":"GET"},"response":{"status":404,"bodySize":128,"content":{"mimeType":"text/css","size":128}},"timings":{"blocked":0,"dns":0,"connect":0,"send":3,"wait":25,"receive":12,"ssl":-1}}]}}

Processing: Normalize each entry, calculate its offset from the earliest startedDateTime, display available blocked/DNS/connect/send/wait/receive/SSL timings, and preserve the aggregate totals and HTTP errors.

Illustrative output:

{ "requests": 2, "totalMilliseconds": 160, "averageMilliseconds": 80, "transferredBodyBytes": 640, "slowest": [ { "url": "https://example.com/api", "method": "GET", "status": 200, "time": 120, "bodyBytes": 512 }, { "url": "https://czoa.com/assets.css", "method": "GET", "status": 404, "time": 40, "bodyBytes": 128 } ], "errors": [ { "url": "https://czoa.com/assets.css", "method": "GET", "status": 404, "time": 40, "bodyBytes": 128 } ], "waterfall": [ { "sourceIndex": 0, "url": "https://example.com/api", "method": "GET", "status": 200, "time": 120, "bodyBytes": 512, "domain": "example.com", "mimeType": "application/json", "startedAt": 1788264000000, "phases": { "blocked": 1, "dns": 10, "connect": 20, "send": 5, "ttfb": 70, "download": 14, "ssl": 5 }, "requestHeaders": [], "responseHeaders": [], "body": "", "startOffsetMilliseconds": 0, "endOffsetMilliseconds": 120 }, { "sourceIndex": 1, "url": "https://czoa.com/assets.css", "method": "GET", "status": 404, "time": 40, "bodyBytes": 128, "domain": "czoa.com", "mimeType": "text/css", "startedAt": 1788264000250, "phases": { "blocked": 0, "dns": 0, "connect": 0, "send": 3, "ttfb": 25, "download": 12, "ssl": null }, "requestHeaders": [], "responseHeaders": [], "body": "", "startOffsetMilliseconds": 250, "endOffsetMilliseconds": 290 } ], "slowRequests": [], "largeRequests": [] }

Addressing limits & protocol boundaries

Local file import streams UTF-8 input and normalizes entries as they arrive; the normalized requests stay in memory for the timeline, filters and body view. Pasted text still uses JSON.parse. Files over 100 MB (100,000,000 bytes) are rejected. Peak memory and performance at the maximum file size or very large entry counts have not been verified. log.entries must be an array of objects; missing phase values appear unavailable. Pattern-based redaction covers common credentials in displayed URLs, headers and bodies but cannot guarantee detection of every secret format. Non-finite status or body size is rejected.

Addressing calculations and subnet models in Offline HAR Network Waterfall Analyzer follow IETF RFC definitions. Ensure subnet masks and CIDR prefixes match your network topology.

Offline network inspection boundary

Address parsing and CIDR calculations in Offline HAR Network Waterfall Analyzer run entirely client-side without probing external IP networks or transmitting routing data.

IETF RFC specifications & IANA registries

IETF Request for Comments (RFC) standards and IANA registry references:

HAR 1.2 historical format reference (not a current W3C recommendation)

Offline HAR Network Waterfall Analyzer verification · Audit standards

How to use it

Paste HAR JSON or open/drop a local HAR file (up to 100 MB).
Review request timing, DNS/connect/TTFB/download phases, and the timeline.
Filter requests, inspect a response body, and find slow or large requests.

Frequently asked questions

What does the HAR analyzer calculate?+

It preserves request counts, total and average time, transferred body bytes, up to five slowest requests, and HTTP errors. It also shows start offsets and available blocked, DNS, connect, send, wait (TTFB), receive (download), and SSL timing values.

Which input and results are available?+

Paste HAR JSON or open/drop a local HAR file up to 100 MB (100,000,000 bytes). The page draws a request timeline, supports domain, MIME-type and status filters, shows slow requests (at least 1,000 ms) and large responses (at least 1 MiB), and lets you inspect a response body. Listed URLs are not fetched.

What are the parsing and privacy limits?+

log.entries must be an array of objects. Local files are read as a UTF-8 stream and each entry is normalized as it arrives; normalized requests remain in memory for the timeline and filters. Pasted text uses JSON.parse. Peak memory/performance at 100 MB (100,000,000 bytes) or very high entry counts has not been verified. Missing phase values appear unavailable. Pattern-based redaction covers common credentials in displayed URLs, headers and bodies, not every possible secret format.

What does the worked example demonstrate?+

The example preserves the aggregate totals for two entries and adds a timeline with available phase values. This bounded example does not establish performance for every 100 MB (100,000,000 bytes) input or prove universal secret detection.