Skip to content

Steel Browser

Set up your agent

Give your coding agent a Steel browser and the Steel skills, so that it browses and writes Steel code the right way. Learn more

Install the Steel CLI. It signs you in and installs the skills for your coding agents.

curl -fsS https://setup.steel.dev | sh

Or send this prompt to your coding agent, and it does the same:

Read https://setup.steel.dev/install.md and follow it to set up Steel for this coding client.
Then continue my requested task using the installed Steel skill.

Set up your agent →

Steel Browser is Steel's open-source browser API. It runs one Chrome browser behind a REST API, on a machine that you control. The code is on GitHub under the Apache 2.0 license.

Steel Cloud is the service at api.steel.dev. Steel runs its browsers for you, adds features that Steel Browser does not have, and bills you for usage. With Steel Browser, you run the browser yourself, and you take care of its capacity, updates and security.

What you get

Steel Browser is a Node.js server that starts Chrome and controls it with Puppeteer over CDP (Chrome DevTools Protocol). It serves its API on port 3000. The Docker image includes Chromium, so you do not need a browser on the host.

For the features that both have, Steel Browser uses the same API paths as Steel Cloud:

  • Sessions, with POST /v1/sessions and options such as proxyUrl, blockAds, dimensions and userAgent. You can list, read and release sessions.
  • Quick actions, POST /v1/scrape, /v1/screenshot and /v1/pdf, which work on a page in one request.
  • Commands, at POST /v1/sessions/{sessionId}/commands, which drive the session. They are Experimental and off by default. See Configuration.
  • Session files, under /v1/sessions/{sessionId}/files, which move files in and out of the browser.

Playwright and Puppeteer connect to a session's websocketUrl, as they do with Steel Cloud. The TypeScript and Python Steel SDKs work with Steel Browser for sessions, scrapes and session files when you change their base URL. See Create a session.

Screenshots and PDFs work differently. Steel Browser's /v1/screenshot and /v1/pdf return the image or the PDF file itself, while Steel Cloud returns JSON with the file's url. The SDK's screenshot and pdf methods expect that JSON, so call these two endpoints with an HTTP client and save the response body. A scrape with screenshot or pdf set returns each file as a base64 string, not as an object with a url.

Steel Browser also has a UI to watch the browser, at /ui in the single Docker image. Its OpenAPI documentation, at /documentation, describes the API of the version that you run.

Steel Browser runs one browser, so it has one session at a time. Creating a session ends the session that is running. Steel Browser does not check an API key, so anyone who can reach its ports can control the browser.

How Steel Browser differs from Steel Cloud

Steel Browser and Steel Cloud are separate code bases, and they change separately. For the API of the version that you run, see /documentation on your instance.

FeatureSteel CloudSteel Browser
Sessions at the same timeDepends on the plan. See Limits.One
AuthenticationAn API key on every requestNone
Screenshot and PDF responsesJSON with the file's urlThe file itself
ProxiesSteel proxies and your own proxyYour own proxy, with proxyUrl
Dedicated IPsYesNo. The browser uses the host's network.
CAPTCHA solvingYesNo
Bot detectionSteel's fingerprints and stealth featuresBasic: stealth plugins and fingerprint injection
CommandsYesOff by default. Turn it on with ENABLE_COMMANDS_API=true.
ExtensionsUpload them with the Extensions APIPut them in the api/extensions folder, build the image, and name them in the extensions option of the session
ProfilesProfiles APINo Profiles API. persist: true keeps one browser data folder.
CredentialsYesNo
FilesThe Files API and session filesSession files only
Live viewYesA UI at /ui
Agent traces, signals, session costYesNo
MCP serverHosted at https://mcp.steel.dev/mcpNo
Updates and capacitySteel handles themYou handle them

When to self-host

Self-host when the browser must run on your own network, for example to reach an application that is not on the internet. The browser then uses your network's DNS, routes and IP address.

Self-host to develop and test against a browser on your own machine, or when you want to change the server itself. Steel Browser is open source, so you can read and change every part of it.

Choose Steel Cloud when you need many sessions at the same time, Steel proxies, CAPTCHA solving, profiles or credentials, or when you do not want to run and secure the servers. Your code changes little between the two, because the Steel SDK takes a base URL and the shared parts have the same paths.

To start, run Steel Browser with Docker on your machine or on a server that you manage.