ToolsArena24

User Agent Parser

Maintained by the ToolsArena24 team. Methodology and test scope

Quick answer: Parse a supplied user-agent string into likely browser, engine, operating system, device and CPU fields using local UAParser rules.

About this user agent parser

Inspect a user-agent string from a request log, browser report or test fixture. The parser uses the locally bundled UAParser rules to identify likely browser name and version, rendering engine, operating system, device model or type, and CPU architecture. Results are shown as a structured JSON object, along with the original user-agent string, making it easier to compare parser output with the actual header that was supplied.

User-agent parsing is pattern matching rather than a direct hardware measurement. Browsers can reduce or freeze their reported details, privacy settings can limit information, and any client can send an invented string. Unknown values may therefore be absent even when the client is perfectly valid. The package version defines the available recognition patterns; future browser releases may need updated rules to be identified precisely.

Use this information for troubleshooting and rough analytics, not as the sole basis for authentication or security controls. Prefer feature detection when deciding whether a web API is available. This tool parses the text you provide and does not contact the device or fetch additional high-entropy client hints.

Which controls should I check first?

The User Agent Parser form labels its controls as Input; Result. Use those labels to keep the source and settings separate. A sample argument that selects a mode is not additional source text. Choose the setting matching your intended operation before copying a result, and change it explicitly when comparing another mode. The labels describe what this form accepts; the worked examples below demonstrate only the stated supplied values.

How should I interpret this specific operation?

The supplied string is a claim made by a client. Parser rules can identify likely browser and operating-system fields, but clients can alter the string. A missing or unknown model should remain unknown rather than being filled in from another person's device. For feature support, checking the actual feature is more reliable than assuming that an identified browser family implements every interface.

Worked helper values

These values come from the reproducible calculation examples supplied with this tool. Each row identifies the complete supplied arguments and the returned value. A visible ↵ denotes a line break and ⇥ denotes a tab in text examples. Settings appear alongside source data so that a different mode is not mistaken for an identical input. These are helper calculations, not observations of a live device, fetched target, image decoder, browser preview or downloaded file. Use the stated method above to relate the returned fields to the actual operation.

Supplied valuesCalculated return value
Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/120.0.0.0 Safari/537.36{"ua": "Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/120.0.0.0 Safari/537.36", "browser": {"name": "Chrome", "version": "120.0.0.0", "major": "120"}, "engine": {"name": "Blink", "version": "120.0.0.0"}, "os": {"name": "Windows", "version": "10"}, "device": {}, "cpu": {"architecture": "amd64"}}

What is supplied data and what is an inference?

The input is a supplied record, string or local file. The output can divide that data into fields or infer likely labels using documented rules. Keep the exact source with the result so that another person can check the interpretation. A parsed field is not automatically a verified fact about the outside world. A user-agent string can be changed, a metadata tag can be edited, and an extracted address can be unavailable. Preserve unknown or missing fields rather than filling them with plausible values that were never present in the source.

Why should I inspect related fields together?

A single label can conceal ambiguity that other fields make visible. A URL has a host, path, query and fragment; a file can have a filename, claimed media type and byte signature; photo metadata can have several different dates. Keep those distinctions in the report rather than collapsing them into one convenient summary. Compare each value with the kind of source it describes. A file extension does not define its actual byte format, and a date stored in an editable tag is not an independent measurement of when the picture was created.

How do I handle absent information?

Treat an absent field as absent. It can mean that the source never contained the information, that its format does not support it, or that the parser did not recognize it. An empty text field can also differ from an omitted field. Read any status message before inferring that the entire input failed. If the result depends on rules or a supported format list, compare a minimal recognized example first. Add a precise note about the missing source value instead of inventing a device model, owner, timestamp or destination based on surrounding clues.

What does the helper example verify?

The example supplies all of its input values explicitly and reports what the local function returned. That demonstrates parsing, normalization or field preparation for those values. It does not verify a device, fetch a website, prove file safety or authenticate metadata. The distinction remains even when an output label resembles a factual statement. A parsed browser family, for instance, comes from the supplied text rather than contacting that browser. Read the example as a reproducible interpretation of data and use separate evidence for claims about the actual external source.

Is the detected device guaranteed correct?

No. The string can be incomplete, frozen or spoofed.

Why are some fields missing?

The local recognition rules did not find enough information.

Should I use this for feature support?

Prefer actual feature detection rather than a browser-name guess.