Browserbase vs Steel
Browserbase and Steel both hand your agent a cloud Chrome over CDP, with proxies, CAPTCHA solving and saved logins. Browserbase is a wider platform with 6-hour sessions, four regions and its own search, fetch and agent APIs, while Steel is an open-source browser you can host yourself, with 1-hour sessions on Scale and browsers in us-east.
The companies and products
Browserbase
Browserbase calls itself the complete platform to build and deploy agents that browse and interact with the web like humans: browsers, search, fetch, functions, model access and identity behind one API key.
Its core product is a Chrome session you drive over CDP with Playwright, Puppeteer or its open source Stagehand SDK, with proxies, CAPTCHA solving, saved browser profiles (Contexts), live view and session replays. Around it sit Search and Fetch APIs (Authenticated Fetch reads logged-in pages without a browser), short TypeScript Functions, Runtime (sandboxed environments that back the hosted Agent shell), a Model Gateway and hosted Agents that work from a plain language goal. It is a browser platform, not a general VM: your code gets the browser, and the Agent gets a shell in a short-lived runtime, not root on a machine. This page compares where the two overlap, which is agents that browse.
Steel
Steel is an API that runs fleets of cloud browsers for AI agents: you create a session, connect Puppeteer, Playwright or Selenium, and release it when done.
Each Steel session is a headful Chromium in a VM in Steel's cloud. You drive it over CDP, a Commands API or a computer-use input API, and you can embed a live viewer. Around it Steel sells managed proxies, CAPTCHA solving, saved browser profiles, a credentials vault that signs in for the agent, a files API and readable traces. It is a different kind of product from boat: a managed browser, not a general sandbox. This page compares where the two overlap: agents that use a browser.
Which use cases each one fits
For each use case, Battleships prices every sandbox provider on the same workload and sorts them from the cheapest monthly bill up. “23rd cheapest of 356” means 22 of the other 355 providers run that workload, with every feature it needs, for less. “With limits” means the provider runs it but misses something, so it is listed after every full match. Open a use case for its whole ranking and what each provider misses.
| Use case and its workload | Browserbase: price rank | Steel: price rank | boat: price rank |
|---|---|---|---|
| Software factory 100 at once · 4 vCPU / 8 GB · 2,200 sessions of 8 h a month Neither hosted browser gives a shell, Docker or a machine disk, so installs, builds and tests need a machine like boat's (or a self-hosted steel-browser next to your own code). | 172nd of 365, with limits $1,255 a month | 157th of 365, with limits $1,115 a month | 5th cheapest of 365 $634 a month |
| Always-on AI employees 50 machines 24/7 · 2 vCPU / 4 GB An assistant that works all month runs into a session cap on both: 6 hours on Browserbase and 1 hour on Steel Scale, so it keeps restarting from a saved Context or profile; boat has no session cap once auto-stop is off (paid plans; default 1 hour). | 286th of 361, with limits $3,699 a month | 128th of 361, with limits $3,070 a month | 6th cheapest of 361 $657 a month |
| Code interpreter bursts 100 at once · 1 vCPU / 2 GB · 200,000 sessions of 1 min a month | 267th of 336, with limits $382 a month | 262nd of 336, with limits $333 a month | 189th of 336, with limits $2,000 a month |
| Sandboxes for your users 300 at once · 2 vCPU / 4 GB · 30,000 sessions of 23 min a month 300 sandboxes at once is more than the 100 browsers Browserbase Startup and Steel Scale allow, and each user gets a browser rather than a machine. | 245th of 349, with limits $1,099 a month | 240th of 349, with limits $990 a month | 3rd cheapest of 349 $209 a month |
| RL rollouts 1,000 at once · 2 vCPU / 4 GB · 1,000,000 sessions of 3 min a month | 259th of 349, with limits $5,049 a month | 252nd of 349, with limits $4,150 a month | 39th of 349, with limits $2,000 a month |
| Agent evals 250 at once · 4 vCPU / 8 GB · 15,000 sessions of 20 min a month | 346th of 359, with limits $449 a month | 345th of 359, with limits $400 a month | 10th cheapest of 359 $200 a month |
| CI runners 40 at once · 8 vCPU / 16 GB · 6,000 sessions of 12 min a month | 245th of 374, with limits $152 a month | 239th of 374, with limits $120 a month | 2nd cheapest of 374 $86 a month |
| 24/7 dev boxes 1 machine 24/7 · 4 vCPU / 8 GB A dev box has to keep its files and processes, and both rebuild the browser for each session, so this fits boat. | 264th of 371, with limits $96 a month | 255th of 371, with limits $73 a month | 19th cheapest of 371 $26 a month |
| Computer-use agents 50 at once · 4 vCPU / 8 GB · 20,000 sessions of 15 min a month Steel has a computer-use input API and human takeover for its browser, Browserbase acts on pages through Stagehand, and boat streams a 60fps desktop where an agent can use any Linux app. | 199th of 367, with limits $416 a month | 190th of 367, with limits $367 a month | 7th cheapest of 367 $200 a month |
| Browser-use agents 50 at once · 2 vCPU / 4 GB · 30,000 sessions of 10 min a month All three fit: Browserbase and Steel give a managed Chrome with proxies and CAPTCHA solving, and boat gives Chromium on a full VM where the same agent can also run code. | 23rd cheapest of 356 $449 a month | 24th cheapest of 356 $470 a month | 7th cheapest of 356 $200 a month |
| Scraping & bots (startup scale) 50 at once · 1 vCPU / 2 GB · 100,000 sessions of 3 min a month Both fit at startup scale; Browserbase also sells Search and Fetch APIs so some pages need no browser, and Steel sells scrape, screenshot and PDF tools per call. | 59th cheapest of 343 $537 a month | 61st cheapest of 343 $540 a month | 6th cheapest of 343 $90 a month |
| Mass scraping & bots (undetectable) 1,000 at once · 1 vCPU / 2 GB · 3,000,000 sessions of 2 min a month Both are built for sites that block bots, with residential proxies and CAPTCHA solving; Browserbase adds Verified browsers on Scale and Steel a stealth browser on Enterprise, while boat uses datacenter IPs. | 12th of 327, with limits $9,799 a month | 8th of 327, with limits $7,950 a month | 73rd of 327, with limits $1,800 a month |
| Parallel red teaming 200 at once · 2 vCPU / 4 GB · 5,000 sessions of 30 min a month | 259th of 351, with limits $291 a month | 250th of 351, with limits $242 a month | 3rd cheapest of 351 $200 a month |
| GPU eval jobs 10 at once · 8 vCPU / 32 GB · 500 sessions of 30 min a month | Not ranked: misses a requirement | Not ranked: misses a requirement | Not ranked: misses a requirement |
| macOS agents 5 at once · 4 vCPU / 8 GB · 1,000 sessions of 20 min a month | Not ranked: misses a requirement | Not ranked: misses a requirement | Not ranked: misses a requirement |
Key features
| Feature | Browserbase | Steel | boat |
|---|---|---|---|
| Kind of product | Managed headless Chrome, plus Search, Fetch, Functions, a Model Gateway and hosted browser Agents | A managed headful Chromium per session over CDP, with proxies, profiles, a credentials vault and a files API | A full Linux VM per sandbox, with a desktop and Chromium inside |
| Isolation | 1 browser per VM in its own subnet; the VM is destroyed after each session | Each session is a Chromium in a VM, per the docs; the VM type is not named | Full VM with its own kernel; vCPUs are shared, not dedicated cores |
| Machine size | Not published; billed per browser hour | Not published; billed per browser hour | 4 vCPU / 8 GB by default; up to 16 vCPU / 32 GB on the $100 plan or higher |
| Shell, Docker and nested virtualization | No shell on the browser VM; hosted Agents and Functions get a shell in a short TypeScript runtime | A hosted session gives browser control only: CDP, REST and a Commands API; you can self-host the open-source steel-browser in Docker next to your own code | SSH, root, systemd, Docker and /dev/kvm on every machine |
| Session length | 15 minutes on Free, 6 hours on Developer and Startup, 6+ hours on Scale | 15 minutes on Launch, 1 hour on Scale, up to 24 hours on Enterprise; longer jobs chain sessions | No cap once auto-stop is off (paid plans); auto-stop defaults to 1 hour and can be set up to 30 days |
| Files after a session | Contexts keep the browser profile (cookies, localStorage, IndexedDB), encrypted at rest | Session uploads and downloads become org-wide Global Files that mount into later sessions; profiles keep cookies and logins (300 MB, deleted after 30 days unused) | The disk is saved every minute and at each stop, and kept at least 14 days after a stop; memory, processes and /tmp are not saved |
| Browser and computer use | Chrome with live view, session replay and logs; Stagehand act, extract and observe | Headful Chrome, a computer-use input API, an embeddable live view with human takeover, mobile mode and traces | 60fps desktop and Chromium, computer use on any app, embeddable stream |
| Bot detection | Residential proxies in 200+ countries, CAPTCHA solving on paid plans, Verified browsers on Scale | Residential proxies, CAPTCHA solving and dedicated IPs on paid plans; a stealth browser on Enterprise | Datacenter IPs and Chromium; no CAPTCHA solving |
| Web APIs beyond the browser | Search and Fetch APIs, Functions that run next to the browser, a Model Gateway and hosted Agents | Browser tools for scrape, screenshot and PDF, billed per 1,000 calls | No hosted search or fetch API; your own code and tools run on the VM |
| Logins and secrets | Contexts keep logins; 1Password for Agents signs in without showing the password; Secrets in Functions; Model Gateway for OpenAI, Anthropic and Google | Credentials API (beta) signs in for the agent without showing it the password | Optional paid AI credentials proxy ($1 a day it is on, plus $1 per 1,000 requests) for model providers and custom HTTPS APIs; other secrets still enter the sandbox |
| Reaching the machine | CDP connect URL and live view links only; no ports to host on | No inbound ports; a CDP websocket and an unauthenticated live view URL; dedicated outbound IPs at $5 an IP a month on Scale | HTTPS URLs for up to 50 ports, token-protected; raw TCP to the sandbox IP (IPv6 only on some sandboxes; the IP changes at each resume) |
| Coding agents | Its own hosted browser Agent from a plain language goal, with built-in tools only | A CLI, skills and an MCP server give your coding agent a Steel browser; the agent runs elsewhere | Claude Code, Codex, OpenCode and more preinstalled, with a prompt, steer and events API |
| Open source and self-hosting | Managed cloud only; the Stagehand SDK is open source under MIT | Open-source steel-browser (Apache-2.0) in Docker, and Steel Outpost inside your own network | Managed cloud only |
| Regions | Oregon, Virginia, Frankfurt and Singapore, picked per session | us-east only; proxy location sets the visible IP; Steel Outpost runs in your own region | Germany, Finland and France; you cannot pick a region; no US or Asia region |
Every feature Battleships tracks (43)
| Feature | Browserbase | Steel | boat |
|---|---|---|---|
| Full VM (own kernel) | Yes | Yes | Yes |
| Docker inside | No documented way | No documented way | Yes |
| Nested virtualization | No documented way | No documented way | Yes |
| Root | No documented way | No documented way | Yes |
| Systemd | No documented way | No documented way | Yes |
| Desktop GUI | No documented way | No documented way | Yes |
| Browser | Yes | Yes | Yes |
| Computer-use API | No documented way | Yes | Yes |
| SSH | No documented way | No documented way | Yes |
| Persistent disk | No documented way | Yes | Yes |
| Pause/resume | No documented way | No documented way | Yes |
| Automatic snapshots | No documented way | Yes | Yes |
| Fork/clone | No documented way | No documented way | Yes |
| Live fork (no pause) | No documented way | No documented way | Yes |
| Memory fork | No documented way | No documented way | No documented way |
| Custom image (Docker or snapshot) | No documented way | No documented way | Yes |
| ≥24 h sessions | No documented way | Yes | Yes |
| Idle auto-stop | Yes | Yes | No documented way |
| Wake on request | No documented way | No documented way | No documented way |
| HTTPS preview URLs | No documented way | No documented way | Yes |
| Raw TCP inbound | No documented way | No documented way | Yes |
| Public IPv4 | No documented way | No documented way | No documented way |
| Egress allowlist | Yes | No documented way | No documented way |
| Secret proxy | Yes | No documented way | Yes |
| Secret proxy for any API | No documented way | No documented way | Yes |
| Preinstalled agents | Yes | No documented way | Yes |
| Agent harness API | No documented way | No documented way | Yes |
| MCP server | Yes | Yes | No documented way |
| GPU | No documented way | No documented way | No documented way |
| Windows | No documented way | No documented way | No documented way |
| Arm64 | No documented way | No documented way | No documented way |
| Self-hosting | No documented way | Yes | No documented way |
| BYOC | No documented way | Yes | No documented way |
| Open source | No documented way | Yes | No documented way |
| Scoped API keys | Not published | Not published | Yes |
| Spend limits | Not published | Not published | Yes |
| Usage per sandbox (API) | Not published | No documented way | Yes |
| Audit logs | Not published | Not published | No documented way |
| SSO | Yes | Yes | No documented way |
| SOC 2 | Yes | No documented way | No documented way |
| HIPAA | Yes | Yes | No documented way |
| EU data residency | Yes | No documented way | No documented way |
| Zero data retention | Yes | Not published | Yes |
Performance
There is no comparable measurement. Browserbase and Steel are not in the public benchmarks we use, and neither publishes the vCPU or RAM behind a browser session, so we cannot say which is faster. For reference, boat did 28.06 runs/s on the V8 Web Tooling Benchmark at 4 vCPU / 8 GB, second of 15 in StarSling's run of 3 October 2026 (run 37073680030). Every boat sandbox in that run ran on an AMD Ryzen 9 9950X host. boat also places some sandboxes on AMD Ryzen 7 9700X hosts, which this run did not measure. Steel says a session starts in under a second; boat is not in the public start-time benchmark.
V8 Web Tooling Benchmark at a 4 vCPU / 8 GB target, runs per second, higher is better. Source: StarSling's run of 3 October 2026 (run 37073680030), median of 5 sandboxes. Providers ran on different hardware in different regions. StarSling gives a shared rank when the run cannot separate two providers. Every boat sandbox in this run ran on an AMD Ryzen 9 9950X host. boat also places some sandboxes on AMD Ryzen 7 9700X hosts, which this run did not measure. Browserbase and Steel are not in the public benchmarks we use, so we cannot say which is faster.
Pricing in five scenarios
Each bill is priced from the providers' public price lists by the open Battleships engine, with the cheapest plan that runs the scenario. Start limits, per-start fees and plan minimums are in the price, boat's included.
| Scenario | Browserbase, a month | Steel, a month | boat, a month |
|---|---|---|---|
| One coding agent, workdays 4 vCPU / 8 GB, 50 GB disk, 8 hours a day, 22 days a month, files kept between days | $99 · Startup · with limits | $250 · Scale · with limits | $20 · $20/mo |
| One agent that never stops 4 vCPU / 8 GB, 50 GB disk, running 24/7 all month | $122 · Startup · with limits | $250 · Scale · with limits | $26 · $20/mo |
| 50 agents in parallel 50 machines of 4 vCPU / 8 GB at once, 8-hour sessions, 1,100 sessions a month, Docker inside | $841 · Startup · with limits | $784 · Scale · with limits | $317 · $20/mo × 3 seats |
| Sandboxes for your users 2 vCPU / 4 GB, 30,000 sessions of 10 minutes a month, up to 300 at once | $499 · Startup · with limits | $510 · Scale · with limits | $200 · $100/mo × 2 seats |
| 1,000 short jobs at once 2 vCPU / 4 GB, 1,000 at once, 3 minutes each, 200,000 jobs a month | $1,049 · Startup · with limits | $950 · Scale · with limits | $8,800 · $2000/mo · with limits |
Browserbase: Plans include browser hours: 100 on Developer ($20 a month) and 500 on Startup ($99 a month), then $0.12 or $0.10 per browser hour, billed by the minute with a one-minute minimum. Proxy data is extra after 1 GB or 5 GB ($12 or $10 per GB), and Search, Fetch and Agents have their own allowances.
Steel: Steel bills browser hours by the minute, rounded up: $0.10 an hour on the free Launch plan, which comes with a one-time $30 credit, and $0.08 on Scale, which costs $250 a month and includes $100 of usage each month. Proxy bandwidth ($10 or $6 per GB) and CAPTCHA solves are billed on top.
Steel: You pick a monthly plan from $20. The plan price comes back as VM time for that month: $20 is about 555 hours of a 4 vCPU / 8 GB sandbox, billed per second. The $20 plan runs 100 sandboxes at once and includes 12 starts a minute, 60 an hour and 200 a day. Create, fork and resume each count as one start; a resume onto the kept machine does not count. Extra starts cost $0.03 to $0.07 each, up to 1,200 a day, and starts above that fail. A stopped sandbox uses no machine time and keeps its disk for at least 14 days. Saved snapshots above the included count and the AI credentials proxy still cost money.
Limits in those scenarios
None of the scenarios fit either one as written, because each prices a browser session with no published size and no machine disk, not a 4 vCPU / 8 GB machine. Session caps bite first: 6 hours on Browserbase Developer and Startup, and 15 minutes on Steel Launch or 1 hour on Steel Scale, so long jobs restart from a saved Context or profile. Both cap self-serve plans at 100 browsers at once, so 300 sandboxes for your users and 1,000 short jobs at once need a sales plan (Browserbase Scale, Steel Enterprise) with no public price. On boat the 1,000-job burst does not fit either: on the $2,000 plan, starts above 90 a minute cost extra, starts above 270 a minute fail with 429, and boat never queues a start. boat auto-stop defaults to 1 hour, and the IP address changes at each resume.
| Scenario | Browserbase | Steel | boat |
|---|---|---|---|
| One coding agent, workdays |
|
| None for this scenario |
| One agent that never stops |
|
| None for this scenario |
| 50 agents in parallel |
|
| None for this scenario |
| Sandboxes for your users |
|
| None for this scenario |
| 1,000 short jobs at once |
|
|
|
Which one should you use?
Use Browserbase when
your agent only needs the web and runs for hours: 6-hour sessions on self-serve plans, Search and Fetch APIs, browsers in the US, Europe or Singapore, and Verified browsers for strict bot walls.
- Longer sessions on self-serve plans. Useful for long browsing jobs: sessions run up to 6 hours on Developer and Startup, where Steel Scale stops at 1 hour.
- More than a browser. Useful when lookups need no browser: Search and Fetch APIs, Functions next to the browser, a Model Gateway and hosted Agents.
- Four regions. Useful if your users or target sites are spread out: browsers run in Oregon, Virginia, Frankfurt or Singapore.
Use Steel when
you want an open-source browser you can host yourself, human takeover and a computer-use API for the browser, its own Credentials API that signs in for the agent, a $0 plan to start, and browsers in us-east.
- Open source and self-hosted. Useful if browsers must run on your own machines: steel-browser is Apache-2.0, and Steel Outpost runs inside your network.
- Human takeover and computer use. Useful when a person must step in: an embeddable live view with takeover, a computer-use input API, mobile mode and traces.
- A vault that signs in. Useful for agents that log in to sites: Steel's own Credentials API signs in without showing the agent the password, with no third-party password manager needed.
Where boat fits
Browserbase and Steel both sell a browser, not a machine. If what you are really running is agents that browse and then also code, build or serve, a boat sandbox puts Chromium on a full VM next to a shell, Docker and a disk that boat keeps at least 14 days after a stop. You can still open a Browserbase or Steel session from boat for sites that block bots.
A whole machine next to the browser
A hosted Browserbase or Steel session gives the agent Chrome over CDP and no shell, so cloning a repo or running Docker needs a second machine (or a self-hosted steel-browser next to your code) and files moved between the two. boat has root, Docker, systemd and /dev/kvm.
No session cap, a disk that stays, and fork
Browserbase stops a session at 6 hours and Steel Scale at 1 hour; a Context, a profile or Steel's Global Files carry over, not the machine. boat has no cap once auto-stop is off (paid plans; default 1 hour), saves the disk every minute and can fork a running sandbox from its last snapshot.
Agents and keys on one machine
Browserbase's Agent has fixed built-in tools, and Steel's CLI needs your agent hosted elsewhere. boat ships Claude Code and Codex you drive from outside, an optional paid key proxy for model providers and custom HTTPS APIs, and HTTPS on 50 ports.
Use boat when your agent browses and also works: installs, builds, Docker and dev servers on a full VM with no session cap once auto-stop is off, a disk saved every minute, a 60fps desktop and coding agents built in. Compare it one to one: boat vs Browserbase and boat vs Steel.
boat's own limits
- Plans start at $20 a month. The plan price comes back as VM time for that month. Unused time does not carry over. Source
- Starts (create, fork, resume): the $20 plan includes 12 a minute, 60 an hour and 200 a day. The $2,000 plan includes 90 a minute, 600 an hour and 2,400 a day. Extra starts cost $0.03 to $0.07. A start fails with 429 above 3 times the included starts per minute, or 6 times per hour or per day. Source
- A stopped sandbox has no compute charge. boat keeps its disk for at least 14 days. A saved snapshot keeps it longer; extra saved snapshots and the AI credentials proxy are paid. Source
- Auto-stop defaults to 1 hour. You can turn it off on a paid plan. The trial allows 2 hours at most. Source
- boat saves the disk every minute and at each stop. Memory, running programs and /tmp are not saved: a resume is like a reboot. A machine failure loses writes after the last good snapshot. Source
- Machines run in Germany, Finland and France only. Snapshot copies can be stored outside the EU. Source
- Public IPv6 and outbound IPv4. No dedicated IPv4. The IP changes at each resume. Source
- Hosts are AMD Ryzen 7 9700X and Ryzen 9 9950X. vCPUs are shared. No GPU. You cannot boot your own Docker image. Source
Questions people ask
Is Browserbase cheaper than Steel?
It depends on the job. One coding agent on workdays comes to about $29 on Browserbase Developer, which splits each 8-hour day into two 6-hour sessions. On Steel it comes to about $18 on Launch if it restarts every 15 minutes from a saved profile, or $250 on Scale for 1-hour sessions. 50 agents in parallel come to $841 on Browserbase against $784 on Steel, though neither fits that scenario as a machine. For 1,000 short jobs at once, both need a sales plan with no public price.
Which is faster, Browserbase or Steel?
There is no comparable measurement. Neither was in StarSling's public CPU benchmark, and neither publishes the vCPU or RAM behind a session. Steel says a session starts in under a second.
Can Browserbase or Steel run Docker?
Not inside a hosted session. Both give you a browser over CDP, not a shell on the VM. Browserbase's hosted Agents and Functions can run shell commands in a short TypeScript runtime with no persistent storage. Steel's open-source steel-browser can run in Docker on your own machine, next to your own code. Every boat sandbox has root, SSH, Docker, systemd and /dev/kvm.
How long can a session run on Browserbase and Steel?
Browserbase allows 15 minutes on Free, 6 hours on Developer and Startup and 6+ hours on Scale. Steel allows 15 minutes on Launch, 1 hour on Scale and 24 hours on Enterprise. On boat, auto-stop defaults to 1 hour; you can set up to 30 days, or turn it off on a paid plan.
Is Steel open source? Can I self-host Browserbase?
The steel-browser runtime is open source under Apache-2.0 and runs in Docker, and Steel Outpost runs Steel inside your own network. Browserbase is a managed cloud, though its Stagehand SDK is open source under MIT.
Which handles bot detection better?
Both have residential proxies and CAPTCHA solving on paid plans. Browserbase lists proxies in 200+ countries and Verified browsers on Scale, while Steel sells dedicated IPs on Scale and a stealth browser on Enterprise. boat uses datacenter IPs and solves no CAPTCHAs.
Try boat
Install the CLI, then run boat new.
7-day free trial