Skip to main content

Raw spark data

Occasionally people ask how to get their hands on raw spark data, usually to create cool bots or fancy tools to do fun things with spark's data. This page explains how you can do just that. 😎


Let's say you have a spark viewer URL:

The simplest way to obtain the raw metadata for this session is to append the ?raw=1 query string to the end of the URL, and send off an HTTP request:


You should get some JSON back that looks like:

"metadata": {
"user": {
"type": 1,
"name": "Luck",
"uniqueId": "c1d60c50-70b5-4722-8057-87767557e50d"
"startTime": 1678186955567
// and so on

You can also do JSONPath filtering by appending & e.g.

"type": 0,
"name": "Bukkit",
"version": "git-Paper-386 (MC: 1.19.3)",
"minecraftVersion": "1.19.3",
"sparkVersion": 2


By default, these endpoints only return the metadata associated with the session. This should be sufficient for most use-cases.

If you really-really want the actual sampler/heapdump data too, firstly, please consider downloading/parsing the raw data yourself. It saves resources at my end! (see below for info)

If this isn't feasible, you can also append &full=true to the URL and get everything back, but be warned, this can potentially return tens of megabytes of JSON!

Raw data

The storage service used by spark is called bytebin.
The specific instance used by spark is:

To obtain the raw data for a given profile (e.g., send a HTTP GET request to the user content endpoint:


If successful, you should receive a response similar to this:

HTTP/2 200
Content-Type: application/x-spark-sampler
Last-Modified: Tue, 07 Mar 2023 13:16:44 GMT
Cache-Control: public, max-age=604800, no-transform, immutable


There are currently two content types used by spark:

  • application/x-spark-sampler - sampler data
  • application/x-spark-heap - heap data

You should then use the sampler or heap schema to parse the raw data, depending on what one you get back.

Protobuf format

The raw data transferred between the spark "client" (the plugin/mod) and the spark viewer is encoded using a format called protobuf (or Protocol Buffers). In order to parse the data, you'll need the protobuf schemas.

There are two versions of the schema (although they can both parse the same data):

Example code

Here is a simple CLI tool written with NodeJS which illustrates how the raw data can be parsed:

(this is actually the code that powers the JSON endpoint above 😛)