
GSD Browser
https://github.com/gsd-build/gsd-browser- Category
- Developer Tools
- Rank
- No. 867Tools index
- Pricing
- Open Source
- Type
- TOOL
- Builder
- gsd-build
- GitHub
- 264 stars
- Latest release
- v0.1.25
- Date
About
Fast browser automation CLI built for AI agents — 63 commands for navigation, screenshots, accessibility, network mocking, visual diffing in one Rust binary.
What it does
GSD Browser controls Chrome or Chromium through a command line backed by an automatically started local daemon. It keeps browser sessions alive across commands, supports isolated named sessions, and offers stable element references, assertions, extraction, recordings, live human control, and structured JSON output.
Why it's ranked here
Its strongest case is unusually broad browser control without requiring a full test framework. Stable snapshot references, explicit assertions, isolated sessions, and JSON output suit agent workflows. The authenticated local viewer adds practical human takeover, annotations, recording, and sensitive-mode controls when automation needs supervision.
What's good
The persistent daemon avoids restarting the browser for every action. Versioned element references provide a deterministic interaction pattern after snapshots. Every command can return JSON. Assertions and batches provide explicit pass or fail behavior. The same tool also handles visual comparisons, network rules, traces, HAR exports, PDFs, saved state, and encrypted credential replay.
Tradeoffs
It targets Chrome and Chromium through CDP, so the supplied text establishes no cross-browser coverage. Prebuilt binaries cover macOS and Linux only. The npm and crates.io packages are not published, leaving installation to a shell installer, release downloads, or a source build. Its large command surface also requires workflow discipline.
How to use it well
Use it for AI agents, CI jobs, and developer automation that benefits from a persistent browser. Snapshot first, act through versioned references, resnapshot after page changes, request JSON, and encode checks as assertions or batches. Use named sessions for parallel work. It does not replace a full browser test framework or provide established cross-browser testing.
Technical notes+
Cargo.toml defines a Rust workspace with common and cli members, release LTO, single codegen unit, and stripped binaries. cli/src/main.rs uses Clap for the command surface and accepts browser, CDP, session, and identity options. common/src/ipc.rs implements length-prefixed messages over Tokio Unix streams with a 16 MB input limit. common/src/lib.rs defines JSON-RPC request and response types plus session-aware daemon paths. common/src/cloud.rs validates cloud input and converts it into the canonical event model in common/src/viewer.rs. common/src/config.rs merges defaults, user and project TOML, environment overrides, then caller-applied CLI flags.
Observed
- License
- Dual licensed under MIT or Apache License 2.0, at the user's option.
- Primary language
- Rust, organized as a Cargo workspace.
- Interface
- Command-line interface backed by a persistent local daemon.
- Browser support
- Controls Chrome or Chromium through the Chrome DevTools Protocol.
- Prebuilt platforms
- macOS on Apple Silicon and Intel, plus Linux on ARM64 and x64.
- Installation
- Shell installer, GitHub release binaries, or Cargo source build.
- Package registries
- The npm and crates.io packages are not published.
- Output surface
- Every command supports structured JSON output.
Read from README.md, Cargo.toml, docs/superpowers/plans/2026-05-02-interactive-browser-workbench.md, cli/src/main.rs, common/src/lib.rs, cli/src/output.rs, common/src/ipc.rs, common/src/cloud.rs, common/src/types.rs, common/src/chrome.rs, common/src/config.rs, common/src/viewer.rs.
What it can do
Automate browser navigation
Navigation commands and target URLs → Automated browser interactions
Take screenshots of web pages
Web page or browser state → Screenshot images
Extract accessibility information
Web page elements → Accessibility data and attributes
Mock network requests and responses
Network request patterns and mock response data → Intercepted and modified network traffic
Compare visual differences between web pages
Two web page screenshots or states → Visual diff report highlighting changes
Execute browser automation commands via CLI
Command line instructions and parameters → Browser automation results and status
Tags
Tech Stack
Comments (0)
No comments yet
Editorially curated, with community endorsements as a secondary signal. Corrections welcome.