steel-browser
🔥 Open Source Browser API for AI Agents & Apps. Steel Browser is a batteries-included browser sandbox that lets you automate the web without worrying about infrastructure.
Runssource: GitHubhomepageTypeScriptApache-2.0commit 2b41124d8e29
TypeScript, Apache-2.0 licensed. The project labels itself: ai, ai agents, ai tools, browser automation, browser use, computer use, headless browser and llm.
steel-browser runs. An Argusic agent installed it in 9.1 minutes and hit 5 errors and fixed every one without help on a clean machine with no GPU, and the whole session was recorded.
With a user-local Node 22 and Chrome for Testing plus DISABLE_CHROME_SANDBOX=true, the dev-mode API runs on port 3000, /v1/health returns 200 with live Chrome, and sessions/scrape/screenshot/pdf plus external CDP client connections all worked with real results.
What the agent ended up with on a clean machine, in its own words. How this is measured
Measured by Argusic on a fresh machine every time. Every number links to its evidence. Argusic Score 100 of 100 (the mean of this project's run scores).
At a glance
Subject data from GitHub, linked at the top of this page, refreshed . Test data by Argusic (CC BY 4.0); every number links to a run page with the full log, the recording, and their sha256 hashes.
What was measured
From the run that decided this verdict. Every line is a recorded value, not an estimate.
What went wrong, in the agent's own words
The agent's own words, worst first, never rewritten into advice.
- No Chrome/Chromium binary present; puppeteer-core does not bundle one2 minutes
- Production-mode API crashed for non-root: FileService hardcodes mkdir '/files' (EACCES); Docker image runs as root so this is a container gap2 minutes
- Container Node was v18 but repo requires >=221.5 minutes
- Chrome launch failed: 'No usable sandbox' on Ubuntu 24.04 (non-root, user namespaces restricted)0.5 minutes
- Stale /tmp/steel-chrome SingletonLock from an earlier killed run blocked browser relaunch0.5 minutes
- Install time
- 9 minutes
- Cold machine to finish
- 11 minutes
- Errors hit and fixed
- 5 hit, 5 fixed with no human help
- How the result was proved
- npm test exited 0 (11 files passed, 79 passed, 2 skipped); curl health returned 200; POST /v1/scrape returned 200 with real page HTML; screenshot 200 decoded by ffmpeg; puppeteer.connect drove a real page navigation
- Model tokens used
- 644,961
- Exact commit tested
- 2b41124d8e29
- Container image
- sha256:cdd920bce783
What was not measured
The other half, so you can see exactly how far this evidence goes. A gap is printed as a gap and never filled in.
- Peak GPU memory
- this ran on a CPU machine, so there was no GPU to sample
- Behaviour over a long run
- each attempt uses the smallest documented settings and a fixed time budget, so this measures whether it runs, not how it holds up under load
Watch the session
The whole run, from a clean machine, at the speed it actually happened. Nothing is cut and nothing is sped up.
A recording of attempt 3, run , the attempt that decided the verdict above. The terminal recording it was made from is the evidence, stored with a sha256 fingerprint; the video is a rendering of it. Nothing loads from the video host until you press play.
The end of the recording
The last lines of the terminal session that decided this verdict, exactly as they were recorded.
3. Production-mode crash for non-root: hardcoded `/files` (EACCES; image runs as root). Worked arou… 4. Chrome sandbox failure on Ubuntu 24.04 as non-root → `DISABLE_CHROME_SANDBOX=true` (documented e… 5. Stale SingletonLock in `/tmp/steel-chrome` from an earlier killed run → removed stale lock files. **Final state:** dev-mode API running on 127.0.0.1:3000 with a live Chrome; health 200; all key API… **Metrics:** - install_succeeded: true - launch_succeeded: true - install_minutes: 9.1 - errors: 5 (all with fixes) - test_depth: tier1_real (real end-to-end: scrape/screenshot/pdf/external CDP client on real pages) - verified_how: `npm test` (11 files, 79 passed, 2 skipped) and curl checks above - final_state: one observed sentence
Replay the whole session, every command from a clean machine to this point.
How it was tested
One agent, one container, no human help. It installs the project the way its own documentation says to, runs it, and fixes what breaks. Everything below is recorded as it happened: the terminal session, the log and the exact commit. The full method.
Strengths and limits
Measured facts, not opinions. How this is written.
What went well
- Reached a running state on a clean machine, with the session recorded.
- Recovered from all 5 errors without a human stepping in, which says the failures are documented well enough to solve.
- Ran without a GPU, so it does not need one to start.
- Licensed Apache-2.0, as reported by its host.
- Measured 3 times, so the result is not a one-off.
What did not, or is not known
- Took 9.1 minutes to install, slower than the median of the 231 comparable projects Argusic has measured.
Of the 232 ai projects Argusic has installed and timed, steel-browser was the 156th fastest to reach a running state, and 153 of 232 reached one at all.
What it is a reasonable choice for
- Trying it on a laptop or a small server: it reached a running state without a GPU.
- A quick evaluation: it was running 9.1 minutes after the clone.
Also tested, in the same area
Every one of these was installed and run by Argusic on a clean machine. Nothing appears here that was not tested.
Run history
Topics (from GitHub)
aiai-agentsai-toolsbrowser-automationbrowser-usecomputer-useheadless-browserllmmcpplaywrightpuppeteer
Embed the badge
Markdown for the project README. It links back here; terms on the terms page.
[](https://argusic.com/subject/steel-browser)Questions
- Does steel-browser run?
- Yes. steel-browser runs. Argusic installed and launched it on a clean machine in 9 minutes, hitting 5 errors on the way, and recorded the session.
- How did Argusic test steel-browser?
- On a fresh, disposable machine, with every command recorded and the repository pinned at commit 2b41124d8e29. 3 attempts are recorded, and the full method is on the methodology page.
- What did it cost to test steel-browser?
- The run that produced this verdict cost $0.14: real compute and model cost, not a list price per million tokens. It is recorded on that run's page.
- How long does steel-browser take to install?
- 9.1 minutes in the run that decided this verdict, from a clean machine to a running state. That is one measured attempt, not an average.
- Does steel-browser need a GPU?
- Not to start. Argusic reached a running state on a machine with no GPU.
- What goes wrong when installing steel-browser?
- 5 things broke in the recorded run, and 5 were fixed without human help. Each one, and the time it cost, is listed on this page.
- How does steel-browser compare with the alternatives?
- Of the 232 ai projects Argusic has installed and timed, steel-browser was the 156th fastest to reach a running state, and 153 of 232 reached one at all.
- Where is the evidence for steel-browser?
- All 3 recorded runs are on this page, each linking to its full log and terminal recording, stored with a sha256 fingerprint so it cannot be quietly altered.