Structured Text Extractor
Extract URLs, IPs, IDs, hashes, tokens, paths and timestamps from logs or raw text
Input
Logs, JSON, CSV or plain text up to 10 MB. Files are read in your browser and never uploaded.
Extracted URLs are never opened.
Findings
Value types
Confidence
- HighChecked by a parser or validator.
- MediumValid shape; the context is ambiguous.
- LowHeuristic match; review it.
Domain detection is a heuristic: endings that look like file names get low confidence.
Email addresses are checked for structure only, not for deliverability.
Results
Related Tools
How to use
- 1.Paste text, import a .log, .txt, .json or .csv file up to 10 MB, or load an example.
- 2.Text up to 200 KB is scanned as you type; for larger input choose Extract to scan it in the background.
- 3.Turn value types on or off, choose how duplicates are grouped, and filter to high confidence.
- 4.Select a value to see its details and choose Find in input to jump to it.
- 5.Export the results as plain lines, JSON, grouped JSON or CSV.
- 6.To work with columns instead, switch to Delimited columns: choose the delimiter and header row, pick columns by number, name or range, then extract or remove them and export CSV, TSV or JSON.
What is the Structured Text Extractor?
The Structured Text Extractor pulls developer-facing values out of logs, pasted payloads, stack traces and other raw text: URLs, domains, email addresses, IPv4 and IPv6 addresses, MAC addresses, UUIDs, ULIDs, JWTs, SemVer versions, Git commit IDs, MD5/SHA hashes, Unix and Windows paths, ISO 8601 timestamps, hex colors, Base64 and Base64url. Every value comes with its line and column, a confidence level and, where useful, a normalized form such as a compressed IPv6 address or an uppercase ULID.
How detection works
Each value type has its own small pattern that finds candidates, and every candidate is then checked by the validator CodingTool already uses for that type: the IP parser rejects 999.999.999.999, the UUID and ULID validators check versions and alphabets, a JWT needs three Base64url parts with a JSON header that names an algorithm, and SemVer uses the same parser as the SemVer Release Advisor. When values overlap, the longer one wins (a URL keeps its domain, a JWT keeps its Base64 parts); Show nested matches lists the inner values too. Domains are a heuristic without a public-suffix list, and email addresses are checked for structure only.
Privacy and limitations
Scanning runs in your browser, large input in a background worker. The text is not uploaded, stored or added to the page address, extracted URLs are never opened, and only your view settings are remembered. JWTs and values the Secret & API Key Scanner engine flags as possible credentials are masked by default, also in exports. CSV exports protect against spreadsheet formula injection. Paths with spaces, compact ISO 8601 forms such as 20261024T101500Z and values longer than 4,096 characters are not extracted.
Frequently asked questions
- Is my log uploaded when I extract values?
- No. Extraction runs entirely in your browser, large input in a background worker. The text and extracted values are not uploaded, stored, logged or added to the URL, extracted links are never opened, and only your view settings such as enabled value types are remembered.
- Why is a value marked low or medium confidence?
- High confidence means a parser or validator confirmed the value, such as the IP parser or the ULID decoder. Medium means the shape is valid but the context is ambiguous, for example 40 hex characters that could be a SHA-1 digest or a Git commit. Low marks heuristic matches such as domain-like file names; turn on High confidence only to hide them.
- Why are some values masked?
- JWTs and values that the Secret & API Key Scanner engine flags as possible credentials, such as database URLs with passwords, are masked by default in the list and in exports. Turn on Reveal values to see or export them, or open the Secret & API Key Scanner to review and redact them.
