Why Steel exists
Steel is an open-source browser infrastructure platform that runs real browsers in the cloud for AI agents. Agents use Steel’s browsers to read pages, click, fill in forms and stay signed in, and developers start each browser with one API call. Hussien Hussien, Steel’s CEO, and Nasr Mohamed, its CTO, co-founded Steel, and they build it with a small team in Toronto. More than 30,000 engineers have built on Steel, which has served more than 35 million browser sessions.
Every AI agent that does real work ends up on a website, and websites are built for people. Logins expire, pages load late, bot checks stop anything that doesn’t look human, and running thousands of browsers at once is a product of its own. Steel builds that product once, so that other teams don’t have to.
Our mission is to build the tools that help AGI serve humanity. We started with the browser, and we want Steel to become the AWS for AGI, a cloud of tools made for agents, which they use to do valuable things for people.
What Steel does
Steel runs browsers in the cloud and gives developers an API to create, control, watch and release them.
Browser sessions in the cloud
One API call starts a real browser in Steel’s cloud, in its own virtual machine that shares no cookies or storage with other sessions, and your code drives it with Playwright, Puppeteer or Steel’s own commands over HTTP. Your agent gets a clean browser for every task, and Launch runs 10 at once, Scale 100 and Enterprise 1,000 or more.
Getting past bot detection
Every session gets a realistic browser fingerprint, and Steel adds residential proxies and CAPTCHA solving for reCAPTCHA v2 and v3, hCaptcha and Cloudflare Turnstile when a site needs them. Your agent is stopped less often on sites that check for automation, and you pay only for what you turn on.
Logins that last
Profiles save a browser’s cookies, logins and settings for the next session, and the Credentials API stores logins encrypted with AES-256-GCM and fills them into a site’s sign-in form. Your agent stays signed in from one run to the next and never needs the password.
Pages as Markdown, screenshots and PDFs
Steel’s Quick actions load a URL in a browser and return the page as Markdown or HTML, a screenshot or a PDF in one request, with no session to create or release. Your agent can hand a page to a language model in one call.
Live view and recordings
Steel shows each running session live, lets a person take control to sign in or solve a CAPTCHA, and records every session by default, with its logs and a timeline of every click and page. You can see what your agent did and find the step where it went wrong.
A CLI, an MCP server and SDKs
Steel has a CLI and skills for coding agents such as Claude Code, Codex and Cursor, a hosted MCP server that chat clients such as Claude connect to, and SDKs for Node.js, Python, Go and Rust. A coding agent can set itself up from one prompt, and your code can start a browser after one install.
How Steel works
An AI agent, a program like Claude Code that does tasks for you, needs a web browser to use a website, the same way you do. Steel gives it one: a real browser that runs in Steel’s cloud instead of on your computer. You add Steel to your agent once, and from then on your agent opens a browser whenever a task needs one, clicks, types and reads the page, and you can watch it work live or play the whole session back afterwards.
What makes Steel different
We win by being obsessed and transparent. The parts of running browsers that stop agents are where we spend our time, and we’re open about how Steel works and what it costs.
We take the hard parts
Bot detection, proxies, CAPTCHAs and logins that last are our problem, so your agent doesn’t stop at the first check. On Enterprise, sessions can run on Stealth Browser, Steel’s own version of Chromium for sites with strict bot detection, at no extra cost.
Your agent doesn’t wait for its browser
A Steel session starts in under a second on average, with no cold-start penalty. In Steel’s Remote Browser Benchmark (November 2025, 5,000 runs for each provider), creating a session, connecting to it, loading a page and releasing it took 0.89 seconds on average on Steel, 1.70 times faster than on Kernel and 1.88 times faster than on Browserbase.
When something breaks, we answer
Scale and Enterprise customers get dedicated support from the people who build Steel. Every plan can write to support@steel.dev or ask on Discord, Scale and Enterprise add a Slack channel shared with the Steel team, and Enterprise adds a 911 email for urgent production incidents.
You see the bill
Every session shows what it cost, and the dashboard splits your spend into browser hours, proxy bandwidth and CAPTCHA solves. Every rate is on the pricing page, and a browser hour costs $0.10 on Launch and $0.08 on Scale, billed by the second.
You can read the code
Steel Browser, Steel’s open-source browser API, is on GitHub under the Apache 2.0 license, so you can see how it works and run it yourself. Steel Browser runs in Docker on your own servers, without a Steel account.
Who uses Steel
More than 30,000 engineers have built on Steel, at companies like Readwise, Cognizant and Profound. Our job is to help the teams who build on Steel succeed and stay, and then to share what works.
Who Steel is for
- AI agent builders who need reliable, signed-in access to the web in production.
- Scraping and data teams who want managed browsers instead of running headless Chrome themselves.
- Automation engineers moving Playwright or Puppeteer scripts to the cloud.
- Teams that keep browser state whose logins, cookies and extensions must last from one run to the next.
What teams build on Steel
Teams build AI agents on Steel, from shopping assistants to customer service agents, and they run web scraping, training data collection for models, RPA software, sales automation and end-to-end tests in real browsers on Steel.
Companies whose engineers build on Steel
- Readwise
- Zo Computer
- Avoca
- Profound
- Octogen
- Clamo
- Atoms
- Melrose
- Fluencify
- Rho
- Provizual
- Zingage
- Cognizant
- Nimbus
- Navattic
- Magical
The people behind Steel
Hussien Hussien, Steel’s CEO, and Nasr Mohamed, its CTO, co-founded Steel. Steel is a small team in Toronto. These are the people who build it, sell it and answer when you write to us.
Hussien HussienCEO
Nasr MohamedCTO
Dane WilsonInfra & Eng
JunHyoung RyuInfra & Eng
Jagadesh PEngineering
Graham LoseeSecurity
Michael SpinaGrowth & Sales
Yeonsu ChoiOperations
Nikola KostadinovicDesign Eng
Key facts
| Name | Steel |
|---|---|
| Company | Nen Labs, Inc. |
| Type | Open-source browser infrastructure platform for AI agents |
| Founders | Hussien Hussien (CEO) and Nasr Mohamed (CTO) |
| Team | 9 Members |
| Headquarters | Toronto |
| Website | steel.dev |
| Core offering | Cloud browsers that AI agents and code control through an API, with stealth, logins that last, live view and recordings |
| Open source | Steel Browser, under the Apache 2.0 license |
| Pricing | Launch is $0 a month plus usage, with $30 of credits once and no credit card. Scale is $250 a month plus usage, with $100 of credits every month. Enterprise is priced by contract. |
| Browser hours | $0.10 an hour on Launch and $0.08 on Scale, billed by the second |
| Plan terms | Launch and Scale are prepaid, and an admin can move between them in the dashboard at any time, effective at once. Enterprise is managed by its contract. |
| Concurrent sessions | 10 on Launch, 100 on Scale and 1,000 or more on Enterprise |
| Support | Email and Discord on every plan, a shared Slack channel on Scale and Enterprise, and a 911 email for urgent incidents on Enterprise |
| Compliance | Servers in the United States, a HIPAA-ready BAA on Scale and Enterprise, and a Data Processing Agreement |
| Users | More than 30,000 engineers, at companies such as Readwise, Cognizant and Profound |
| Served | More than 35 million browser sessions and 1.5 million browser hours |
| Partners | A launch partner of Stripe Projects since July 2026, and a native browser provider in Nous Research’s Hermes Agent |
| Social | GitHub, X, LinkedIn and Discord |
Frequently asked questions
Steel is an open-source browser infrastructure platform that runs real browsers in the cloud for AI agents. An agent or a program starts a Steel browser with one API call and controls it with Playwright, Puppeteer, the Steel CLI or Steel’s MCP server. More than 30,000 engineers have built on Steel.
Steel’s Launch plan costs $0 a month plus usage and comes with $30 of credits once, with no credit card needed. Scale costs $250 a month plus usage, with $100 of credits every month, and Enterprise is priced by contract. A browser hour costs $0.10 on Launch and $0.08 on Scale, and every rate is on Steel’s pricing page.
Steel publishes its browser API, Steel Browser, as open source under the Apache 2.0 license, so you can run it on your own servers, while Browserbase is a managed platform. In Steel’s Remote Browser Benchmark of November 2025, creating a session, connecting to it, loading a page and releasing it took 0.89 seconds on average on Steel, 1.88 times faster than on Browserbase. Steel’s comparison of the two also covers how each keeps logins and how each charges.
Steel was co-founded by Hussien Hussien, its CEO, and Nasr Mohamed, its CTO. Steel is built by a small team in Toronto, and Steel is a product of Nen Labs, Inc.
Steel offers cloud browser sessions through an API, with residential proxies, CAPTCHA solving and realistic fingerprints for sites that block automation, and profiles and stored credentials that keep an agent signed in. Steel also returns a page as Markdown, a screenshot or a PDF in one request, records every session by default, and gives coding agents a CLI, skills and a hosted MCP server.
Yes. Steel Browser, Steel’s open-source browser API, runs with Docker on your own machine or servers without a Steel account, and its code is on GitHub under the Apache 2.0 license. Steel Cloud runs the browsers for you and adds what Steel Browser doesn’t have, such as many sessions at once, Steel’s proxies, CAPTCHA solving, profiles and credentials.
Build something great with us
Each team at Steel owns an outcome rather than a part of the product, and everyone writes their own issues: the problem, the smallest useful release and how we’ll know it worked.
We’re hiring in Toronto. If you want to work on the browsers that AI agents run on, these are the roles open now.