Skip to content

HAR file analyzer and secret sanitizer

Diagnose slow requests and failing API calls locally, then strip session cookies, JWTs, and API secrets from .har captures before attaching them to support tickets.

Load HAR capture (.har / .json)

Drop a browser .har capture here

Processed offline in a local Web Worker. Captured cookies and tokens stay on your device.

Or paste raw HAR JSON

How to use HAR file analyzer

  1. Load a capture or sample trace. Click Load HAR capture (.har / .json), drop an exported browser trace onto the DropZone, paste raw JSON under Or paste raw HAR JSON and click Analyze pasted HAR, or click Load sample HAR.
  2. Isolate slow or failing calls. Filter the waterfall with the All, 2xx, 3xx, 4xx, 5xx, or Has secrets chips, type a host or endpoint into the search box, and open any row via Details.
  3. Choose sanitization toggles. Review the five checkboxes under Sanitize HAR before sharing to redact auth headers, cookies, secret query keys, body credentials, and optional internal hostnames.
  4. Export clean artifacts for support. Click Download sanitized .har to save sanitized-capture.har for vendor engineers, or click Export CSV summary to export har-requests.csv with spreadsheet formula protection.

Everything runs in this browser tab. Nothing you paste or open is uploaded or logged, so it's safe for keys, tokens and private data.

How does offline HAR capture analysis protect session secrets?

Processing HTTP archive logs locally inside browser memory prevents live session cookies, Bearer JWTs, and API credentials from ever crossing external networks. Every validation pass, timing summary, secret scan, and redaction write executes directly on your device without server storage.

Troubleshooting production login redirects or failing API gateways often generates multi-megabyte captures governed by the W3C Web Performance HAR 1.2 specification. Uploading those files to cloud viewers exposes both active user sessions and internal service payloads before engineering triage even begins. When login incidents involve single sign-on redirects, pairing this HAR sanitizer with SAML Decoder keeps both network traces and XML assertions strictly on your machine.

Local execution eliminates three common capture leaks before files leave your laptop:

  • No upload telemetry: Capture text, filenames, and entry counts never trigger fetch, XMLHttpRequest, or analytics payloads.
  • Immediate secret flagging: Red badges mark rows carrying credentials the moment parsing completes.
  • Reproducible redaction: Generating sanitized-capture.har clones the original structure cleanly with predictable placeholders.

Waterfall timings, status filters, and failing requests

Once loaded, four top summary cards report total request count, cumulative transfer bytes plus wall duration, capture creator metadata, and flagged secret entries. Engineers diagnosing backend contract mismatches can also inspect OpenAPI endpoint definitions locally alongside failing routes.

DNS, TLS, and server wait timing breakdown

Each row parses timings phases into total request time and server TTFB wait time (wait), highlighting slow upstream responses immediately. In the built-in sample capture, POST /v1/oauth/token completes in 184.5 ms with 112.5 ms wait, whereas GET /v1/orders?limit=25 fails with HTTP 502 after 742.0 ms total and 731 ms waiting.

Selecting Details opens the full right-hand request inspector showing copied request URLs, complete request and response header tables, and payload previews. Reviewing credential errors under IETF RFC 6750 Bearer Token Usage helps distinguish missing tokens from upstream gateway timeouts.

Filtering by status class, host, method, and secrets

Status filter chips slice the waterfall instantly across All, 2xx, 3xx, 4xx, 5xx, and Has secrets. Combining 4xx or 5xx with the text filter matches substrings across URLs, HTTP methods, hostnames, and MIME types in one pass.

Redacting auth headers, cookies, and secret query params

Support tickets rarely need raw session credentials when debugging HTTP status codes or slow payload transfers. Four redaction checkboxes under Sanitize HAR before sharing turn on automatically so credentials matching IETF RFC 6265 HTTP State Management Mechanism and API key patterns are scrubbed before download.

Sanitizer toggle Default Scope inspected Replacement marker
Auth & API headers Enabled 15 headers including authorization, x-api-key [REDACTED]
Cookies & Set-Cookie Enabled Cookie headers plus request.cookies and response.cookies [REDACTED]
Secret query params Enabled 20 query keys such as token, code, sig, samlresponse REDACTED
Body tokens & passwords Enabled 13 JSON/form keys plus regex patterns in body text [REDACTED_JWT] or [REDACTED_KEY]
Pseudonymize hostnames Disabled URL hostnames, Host headers, serverIPAddress host-N.redacted.invalid, 0.0.0.0

Header sanitization targets 15 known credential fields case-insensitively, covering authorization, proxy-authorization, cookie, set-cookie, x-api-key, x-auth-token, x-csrf-token, and x-amz-security-token. Query string scrubbing inspects 20 parameter keys including access_token, id_token, refresh_token, x-amz-signature, and samlrequest inside both parsed queryString arrays and raw request URLs. When debugging cross-system routing alongside mail relays, HTTP and mail header troubleshooting diagnostics cover adjacent header parsing workflows.

Regex scrubbing for JWTs, Bearer tokens, and API prefixes

Payload bodies often carry secrets nested inside raw JSON, URL-encoded forms, or error text. Thirteen sensitive body field names like password, client_secret, ssn, and credit_card are redacted structurally alongside regex scans for three-segment JWT strings (eyJ...), Bearer or Basic headers, and prefixed vendor keys (sk_, pk_, ghp_, xoxb-, AKIA).

Exporting sanitized HAR captures and CSV request logs

After verifying flagged entries, engineers can download a clean JSON archive or export metrics for incident postmortems. Reviewing stripped OAuth flows works smoothly alongside checking Bearer JWT claims without leaking session tokens in separate local inspection tabs.

Pseudonymizing internal hostnames and server IPs

Enabling Pseudonymize hostnames protects private VPC DNS names, staging subdomains, and internal IPv4 or IPv6 endpoints before sharing traces externally. Every unique hostname receives a consistent numbered alias such as host-1.redacted.invalid, keeping multi-domain call patterns traceable while masking infrastructure topology and clearing serverIPAddress to 0.0.0.0.

Exporting sanitized JSON captures and CWE-1236 safe CSV

Clicking Download sanitized .har serializes the redacted archive with creator metadata intact, while Export CSV summary writes all 9 columns including method, status, duration, timing wait, and transfer size to har-requests.csv. Every CSV cell passes through MITRE CWE-1236 CSV formula neutralization before download.

The CSV exporter prepends a UTF-8 byte order mark, uses CRLF line endings for Excel compatibility, and neutralizes cells starting with equals, plus, at, minus, or pipe characters to prevent spreadsheet formula execution.

What limits apply to large browser HAR captures and previews?

Memory-only browser parsing handles multi-megabyte HTTP archives cleanly, though three deliberate display limits keep browser scrolling responsive on large recordings. Exporting sanitized archives and CSV summaries bypasses UI display limits and always processes every recorded entry.

For smooth rendering, the waterfall table displays the first 300 matching rows after filtering, while domain aggregation summarizes the top 15 hostnames by hit count. Individual request inspection truncates response body previews at 2,000 characters so multi-megabyte assets never stall the UI.

Captures exceeding 50 MB or containing embedded media blobs perform best when filtered in DevTools before exporting or stripped of heavy binary payloads:

  • Full-archive sanitization: Downloading sanitized-capture.har processes every recorded row rather than stopping at the 300-row table limit.
  • Complete CSV rows: Exporting har-requests.csv writes all matching and unfiltered entries with status codes, wait times in ms, and 0 B or actual transfer sizes preserved.
  • Local tab responsiveness: Running this HAR file analyzer inside your browser avoids upload timeouts while protecting multi-megabyte session traces on disk.

Questions

Why should you sanitize a HAR file before sharing with support?

Browser DevTools HAR exports capture live session cookies, Bearer tokens, OAuth authorization codes, and POST passwords in plain text. Anyone who receives an unredacted archive can replay active sessions or inspect customer PII without knowing your account password.

Which headers and tokens does HAR Analyzer redact by default?

Four of five sanitizer toggles start enabled by default. They scrub 15 auth and API headers, Cookie and Set-Cookie values, 20 sensitive query keys, 13 body credential keys, and regex matches for JWTs, Bearer strings, and common API prefixes.

How many requests can the HAR waterfall table display?

The interactive waterfall displays up to 300 filtered rows in the browser UI so large multi-thousand-request traces stay responsive. Sanitization and CSV export still process every entry in the loaded archive without truncating rows beyond row 300.

Does HAR sanitization preserve request timings and status codes?

Yes. Redaction clones the parsed JSON structure and replaces only credential values with labeled placeholders. HTTP status codes, per-phase timing breakdowns, byte transfer sizes, methods, and entry timestamps remain intact for root-cause debugging.

How does hostname pseudonymization alter capture URLs?

Enabling Pseudonymize hostnames maps every distinct origin hostname to a deterministic placeholder like host-1.redacted.invalid across URLs and host headers while setting serverIPAddress to 0.0.0.0. Internal staging topology stays hidden while cross-request domain relationships stay legible.

Can you inspect a HAR trace when no secrets are flagged?

Yes. Read-only traffic without auth headers, cookies, sensitive query parameters, or token-like strings shows zero Sensitive Entries and still populates every waterfall column, timing breakdown, header table, and CSV export column.

Updated