Skip to main content
WhatsApp Web is shipped to your browser as tens of thousands of scrambled JavaScript files. Everything the app can do is in there, including parts that are built but not switched on yet. cellar downloads a version of that code, unpacks it into one readable file per module, and lets you search it, compare two versions, and see which module uses which. Nothing is uploaded anywhere. It reads the same files your browser already downloads.

Set up your AI assistant first

The fastest way to use cellar is to let your coding assistant drive it. One command installs the tool, connects it, and teaches it what to do.

​
Install

You need a recent stable Rust toolchain and just, which runs cellar’s install recipes.
1

Install just

More options, including Fedora, Nix, Scoop and a prebuilt binary, are in just’s install guide.
2

Build cellar

This puts cellar at ~/.cargo/bin/cellar. Make sure ~/.cargo/bin is on your PATH.
Then download a version of WhatsApp Web. This pulls a few hundred megabytes and takes a few minutes. Each version uses about 1.3 GB on disk.
Old versions stay available after WhatsApp moves on, so you can go back and pick up a release you missed and compare against it.

​
What you can do

Use it from an AI assistant

Sixteen tools over MCP, plus a skill that teaches the assistant when to use each.

Follow a real example

Four commands that uncover passkey device linking before it was announced.

Manage versions

Download, import, inspect and remove versions of the client.

Search the code

Find modules by name, by the code inside them, or by what they export.

Compare releases

See what changed, as JSON, NDJSON, Markdown or text.

Follow the wiring

Dependency graphs as Mermaid, Graphviz or JSON.

​
Every command speaks JSON

Anything that produces data takes --json, so cellar composes with jq, scripts and assistants. Comparisons additionally offer NDJSON and Markdown, and graphs offer Mermaid and Graphviz. See comparing releases and graphs. cellar is an independent tool. It is not affiliated with, endorsed by, or sponsored by WhatsApp LLC or Meta. It reads publicly served client code for interoperability research. “WhatsApp” is a trademark of its respective owner and is used here only to identify the software this tool reads.