Electron release diff for security review: what changed in IPC, contextBridge, windows and hosts.
io.github.maximilianfeix/hexvero (MCP Server)
The io.github.maximilianfeix/hexvero MCP server provides local, execution-free analysis of binaries, application bundles, and archives. It is intended for use with a CLI and for agents, focusing on extracting information without running the analyzed files.
🛠️ Key Features
Local analysis
Execution-free processing
Targets binaries, app bundles, and archives
🚀 Use Cases
Command-line interface (CLI) analysis workflows
Agent-driven analysis of binaries and archives
⚡ Developer Benefits
Works with CLI and agent tooling
Avoids executing analyzed files by design
⚠️ Limitations
No additional capabilities beyond “local, execution-free analysis” are described in the provided data
Diff two builds of an Electron app and see the IPC channels, preload bridge, window settings and network hosts that changed, with file and line on both sides.
30 seconds of real output: npx hexvero demo lists what changed between two builds of an example Electron app, including a new preload member with its preload and handler lines and an IPC channel reported as unknown; hexvero check fails the release policy with exit 3; then the local web UI opens the new member and the changed window with the source lines of both builds
Needs Node.js 22.12 or newer. It copies two builds of a small example app (NotesDesk 1.4.0 and 1.5.0) into ./hexvero-demo and shows what changed. Excerpt, … marks cuts:
text
What changed between notesdesk-1.4.0.asar and notesdesk-1.5.0.asar
Two builds of a small example Electron app, read statically. Nothing was executed.
interfaces (2)
+ preload.js::notesdesk::checkUpdate @ b: preload.js:9, handler main.js:30 · link direct
- preload.js::notesdesk::openInShell @ a: preload.js:9, handler main.js:30 · link direct
…
ipcChannels (2)
+ update:check @ b: receiver main.js:30, sender preload.js:9
? shell:open @ a: receiver main.js:30, sender preload.js:9
the other side has values there that could not be resolved
windows (1)
~ main.js::win @ b: main.js:8
nodeIntegration: true → false, nodeIntegrationInEffect: "on" → "off", sandbox: false → true
…
endpoints (1)
+ main.js::fetch::https://updates.notesdesk.example/latest.json @ b: main.js:30 · literal
? means hexvero could not tell, so shell:open is reported as unknown, not as removed. Then turn it into a gate:
text
$ npx hexvero check hexvero-demo/notesdesk-1.4.0.asar hexvero-demo/notesdesk-1.5.0.asar --policy hexvero-demo/policy.json
hexvero check fail: the policy is violated
✗ new-interface-member notesdesk.checkUpdate new · fail
✗ new-external-host updates.notesdesk.example new · fail
…
$ echo $?
3
npx hexvero ui --demo --open shows the same comparison in a local web UI with the source lines of both builds side by side. The full tour: walkthrough.
Why
A release review asks what the update changed: which channels a page can now reach, which preload members appeared, which window got more permissive, which servers the code now calls.
A diff of the shipped bundle does not answer that. It is minified JavaScript inside an app.asar; running the app to find out is not always allowed, and a checklist scanner looks at one build, not the change.
hexvero answers it from the files: each change with the file and line it was read from in both builds, what it could not establish, and a policy that fails CI on a new member, host or more permissive window. Nothing is executed.
What it reads
IPC and bridge map
window → contextBridge member → IPC channel → ipcMain handler, with how each link was read
Window settings
webPreferences per window, and whether nodeIntegration and sandbox are in effect
Runtime and fuses
Electron, Chromium and Node versions, fuses and the ASAR integrity record, read from packaged binaries
Hosts and storage
network endpoints and localStorage/sessionStorage keys, literal or inferred
16 tools with small paged answers, so a coding agent can run the same review
Local web UI
the comparison with both builds' source lines side by side, plus an offline HTML report
Inputs: an app.asar, a macOS .app, a Windows or Linux install directory, or the release .zip you publish. Installers (.dmg, setup .exe, AppImage) need unpacking first. Per-platform paths: docs/cli.md.
On your own app
bash
npx hexvero compare old/YourApp.app new/YourApp.app --report report.html # one offline HTML page
npx hexvero init old/YourApp.app # starting policy, current hosts allowed
npx hexvero check old/YourApp.app new/YourApp.app --policy .hexvero/policy.json
To accept what is there today, add --baseline .hexvero/baseline.json --update-baseline once; later runs fail only on new findings.
One build instead of two, with the single-build checks known from Electronegativity (CSP, shell.openExternal, permission handlers, window settings, certificate and command-line switches; from 0.13 also <webview>, custom protocols, executeJavaScript with code that is not a literal, resources over plain http and an Electron runtime past its end of support):
bash
npx hexvero audit YourApp.app --format sarif > audit.sarif # exit 3 when a rule is violated
Use it in CI
Pinned to a version so a release cannot change your results unannounced:
yaml
jobs:hexvero:runs-on:ubuntu-latestpermissions:contents:readsteps:-uses:actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1# v7with:persist-credentials:false-uses:actions/setup-node@949feb2413d6458794dcd2491c4babbbce0c15c1# v7with:node-version:24# Build your app and fetch the previous release here, e.g. into previous/ and dist/.-run:npx--yeshexvero@0.12.0checkprevious/app.asardist/app.asar--policy.hexvero/policy.json
Baseline, Markdown job summary and a least-privilege SARIF upload for code scanning: docs/ci.md.
For AI agents
In Claude Code, server and skill come as one plugin:
Elsewhere, npx hexvero setup --apply registers the MCP server with the coding agents it finds (dry run without --apply), or run npx -y hexvero@0.12.0 mcp over stdio. Text inside the target is treated as data, never as instructions. Clients, tools and what an agent receives: docs/mcp.md.
Electronegativity was archived on 2026-10-24. From 0.13, hexvero audit <build> runs its security checks with the same exit codes and SARIF as check, except two for options Electron has removed (affinity, auxclick) and its 27 API-upgrade checks; the check-by-check list is in the migration page. REA does much that hexvero does not (dynamic analysis, devices, decompilers); use it for that. Sources, dates and every caveat: docs/comparison.md.
Accuracy and limits
On the newest blind set, written by someone without access to hexvero, it found 52 of 64 IPC channels with none false, and it misses and misreports rows in other lists (all numbers). On real releases it is checked against the source (GitHub Desktop 3.6.7 → 3.6.10, Electron Fiddle 0.38 → 0.39), misses included. It reads files and never runs them, so a finding is what the code contains, not what it does at run time or whether it is exploitable; a passing check is not a security verdict. Parsing a hostile file still processes attacker-controlled bytes, within size, depth and time limits. The full list: docs/accuracy.md.
No. Targets are parsed as bytes; nothing is executed. The UI binds to 127.0.0.1 with a per-session token. There is no telemetry and no network call in the code base. Through a cloud-hosted agent, the results that agent requests go to its model provider like any other tool output. Use hexvero only on software you may examine (DISCLAIMER.md).
Contributing
bash
npm install && npm run check # typecheck, lint, format, unit tests, shape rules
npm run e2e # the user flows on real inputs
Rules in CONTRIBUTING.md; good first issue is sized for an afternoon. Found a wrong or missing fact? Open a bug report with hexvero doctor output and, if you can, the file.
If hexvero showed you something you would have missed in a release, a ⭐ helps the next Electron team find it.