No more memory surprises.

Git tracks your code. Track what it does to your flash and RAM, on every commit and pull request, across all your targets.

Free for open source

Which PR pushed flash from 77% to 93% last month?One board, one point per commit
70%80%90%100%Oct 6Oct 11Oct 16Oct 21Oct 26Oct 31Nov 575%92%

Used by

The later you catch memory bloat, the more it costs

  1. In the pull requestThe author fixes it before merging
  2. When flash runs outNobody saw it coming. Builds fail, and everyone who hits the limit has to stop and find space they didn’t use
  3. Before a releaseNo time left to find the space, so features get cut and the release date slips
  4. After launchUpdates no longer fit on devices in the field, and fixing that can mean new hardware

Reviewers can't see memory

A reviewer reading the diff can't see what it does to flash and RAM. MemBrowse shows the bytes on the pull request and on a dashboard, for every board, before you merge.

Real numbers from Meshtastic PR #12113
meshtastic-firmwaredf9051cc...Showing 4 of 17 targets
Real numbers from Meshtastic PR #12113 · See all 17 targets

Know where your memory went

Click the jump in the chart and you'll find exactly which commit caused the issue. You can load your earlier commits, so you don't have to wait weeks for it to be useful.Tap the jump to find the commit that caused it. You can load your earlier commits, so it's useful from day one.

Memory Region Changes

Drag across the chart to zoom in

Real data from the Flipper One firmware. That spike is a linker change that added 128 KB on June 8, fixed the next day.

See what's inside every change

Compare any two builds to see what changed, then open the symbols to find what's using the memory. You'll know what to trim instead of guessing. Real data from Apache NuttX.Compare two builds to see what changed, then open the symbols to see what's using the memory. Real data from Apache NuttX.

Memory Budgets

Battery monitor

FW-412

Memory Regions

Region
Usage
Limit
Status
FLASH
18.4 KB
32 KB
57.5%

Firmware update screen

FW-538

Memory Regions

Region
Usage
Limit
Status
FLASH
58.6 KB
64 KB
91.6%

TLS library update

FW-611

Memory Regions

Region
Usage
Limit
Status
FLASH
81.92 KB
64 KB
128.0%
RAM
4 KB
8 KB
50.0%

Know when a feature goes over budget

Budget memory for each feature when you plan it, then compare planned against actual as the commits land. You'll know a feature is running over while there's still time to optimize or change hardware.Set a memory budget for each feature and track it as commits land, while there's still time to optimize or change hardware.

Go over budget and CI fails the pull request.

How budgets work
How budgets work

Set up in minutes. Your code never leaves CI.

Setup is one short CI workflow on GitHub or GitLab, and on GitHub our Claude Code plugin can do all of it for you: it finds your ELF files, linker scripts and, optionally, map files, then sets up the workflows and the PR comment.

The CLI, the GitHub Action and the GitLab CI component are open source. In CI, the CLI reads your ELF and linker scripts, and only JSON reports leave CI. If your security policies are stricter, you can run all of MemBrowse on your own servers, so even those reports stay inside your network.

Support answers within 24 hours.

One short CI workflow on GitHub or GitLab, or let our Claude Code plugin set it all up. The CLI, Action and GitLab component are open source, and only JSON reports leave CI.

Stricter policies? Run all of MemBrowse on your own servers.

Support answers within 24 hours.

.github/workflows/membrowse_report.yml
name: MemBrowse Memory Report

on:
  pull_request:
  push:
    branches: [main]

permissions:
  contents: read
  pull-requests: write

jobs:
  analyze:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v5
        with:
          fetch-depth: 2                 # parent commit, for the diff

      - name: Build
        run: make                        # your build command

      - uses: membrowse/membrowse-action@v1
        with:
          target_name: stm32f4
          elf: build/firmware.elf
          ld: linker.ld                  # your linker script(s)
          map_file: build/firmware.map   # optional
          api_key: ${{ secrets.MEMBROWSE_API_KEY }}

      - uses: membrowse/membrowse-action/comment-action@v1
        if: github.event_name == 'pull_request'
        with:
          api_key: ${{ secrets.MEMBROWSE_API_KEY }}
          commit: ${{ github.event.pull_request.head.sha }}
        env:
          GH_TOKEN: ${{ secrets.GITHUB_TOKEN }}
Or let Claude Code set it up
/plugin marketplace add membrowse/membrowse-action
/plugin install membrowse@membrowse-marketplace
/membrowse-integrate

Frequently Asked Questions

Can I build this myself?

You can, and some teams do. Homegrown tools often work well at first, then get surprisingly annoying to maintain. On Meshtastic: “We’ve been trying to hand-roll solutions to this for far too long, and they just kept breaking.”

What makes it harder than it looks?

A script that prints section sizes takes an afternoon, and AI can write it for you. A tool your team can rely on for years, across every target and thousands of commits, is a different project. Toolchain updates break the parser, new projects don’t fit, queries slow down as the data grows, and gaps appear in the history. Nobody was hired to deal with this, so it gets done between other tasks, until it stops getting done. That’s how you end up with a PR comment and no history, or a dashboard that’s out of date and nobody trusts. MemBrowse is battle-tested on many projects and stores over 20 million reports.

How long does it take to move to MemBrowse?

Setup is one short CI workflow. You can load your earlier commits, so the dashboard opens with data in it.

Which CI platforms work?

GitHub Actions and GitLab CI have ready-to-use workflows. Other systems, like Jenkins or CircleCI, can use the CLI directly.

What happens to my data?

MemBrowse stores memory analysis results, never your source code or binaries. Only JSON reports with symbol names, addresses and sizes leave CI.

We have strict security policies. Can we run MemBrowse on-premise?

Yes, and nothing leaves your network. MemBrowse runs entirely on your own servers, and your source code and binaries never leave CI. Get a demo to talk through your requirements.

Free for open source. Paid plans start from $99/month billed annually, with a 15-day free trial. See pricing

Start tracking your firmware
Supported by
Technology Partner
Technology Partner
Listed in vendor offerings