Repository navigation
two browsers in one session
A session here is not one browser. It is two, named main and support, and
every tool takes a browser argument to say which one it means. Most tasks never
need the second. The ones that do need it badly, and the behaviour has one trap
worth knowing before you meet it.
Measured on 2026-09-17, opening each in turn and asking the session what it had:
browser_open -> main opens. identity: seed 982240811
browser_open browser: support -> support opens. identity: seed 1206782028
{"focus": "support",
"browsers": [{"id": "main", "url": "", "urls": []},
{"id": "support", "url": "", "urls": []}],
"note": "2 of 2 browsers. Commands that name none go to main."}Three facts fall out of that, and the third is the trap.
The identities are different. Two seeds, two fingerprints. The second browser is a different person, not a second window belonging to the first. That is the whole reason the feature exists.
The exit is the same. Both go out from the same address unless you say otherwise. The support browser's own opening line says so: same exit as main. Different identity, same address, which is a combination worth understanding before you rely on it.
Focus and default are not the same thing. After opening the second browser
the focus is support, and yet the note says commands that name none go to
main. Both are true and they point in opposite directions. If you open a
support browser and then issue a command without naming a browser, it goes to
main, not to the one you just opened and are presumably looking at.
That asymmetry is deliberate rather than a bug: an unnamed command having a fixed destination is more predictable than one that follows whatever you touched last. It is still the thing that will produce one confusing run before you internalise it. Name the browser on every call once you have two.
Keeping a reference open. Read a specification, a spreadsheet view or a list in support while filling a form in main, without navigating away from a page that holds state. Losing a half-filled wizard by navigating away to check something is a real and annoying failure, and this is the answer to it.
Two accounts at once. Two sessions on the same service, side by side, comparing what each one sees. With separate profiles, two logins that do not interfere.
A clean second opinion. A page behaves oddly in main, which has been clicking around for twenty steps and has accumulated cookies and state. Open it fresh in support and see whether it still does. This is the browser-side version of the bisection in browser problem or model problem.
Comparing two versions of a page. Two markets, two variants, two currencies, read at the same moment rather than two minutes apart. Give each browser its own exit and it becomes the honest way to run the comparison in seeing a page as it appears in another country.
Measured on the same machine: the server alone is 77 MB across two processes, one browser takes it to 704 MB across ten, and the second takes it to 1,261 MB across eighteen. The second browser is 556 MB and eight more processes, a 79% increase on the first.
So it fits comfortably on a 4 GB machine and is most of a 2 GB one before the pages have anything heavy on them. Close it when the comparison is done rather than leaving it open for the rest of the conversation. What a second browser costs has the breakdown and the method for measuring it on yours.
Worth being precise, because "a second browser" sounds like more isolation than it is.
Not the process. Both browsers live in one server process. If that dies, both go.
Not the exit, unless you ask. They share an address by default. Pass a
different proxy to browser_open for the one that needs a different one.
Not your machine. Two identities from one address, browsing at the same time, is a pattern rather than two unrelated visitors. Whether that matters depends entirely on what you are doing, and pretending otherwise would be the kind of absolute claim this wiki avoids.
Not the model's attention. The agent has one context. Two browsers means two sets of page state for it to keep straight, and the most common failure is not technical: it reads support and acts on main.
The tools give you the state; the task has to make the agent check it.
- Say which browser each step is for, in the task, by name.
-
Have it call
browser_listbefore acting on a browser it has not touched for a while. That returns the focus, both ids and the current URL of each, which is enough to catch a mix-up before it becomes a click. - Expect a clear error rather than a silent wrong action when a browser has no page: a tool call against an empty browser answers "this browser has no page open; browser_navigate opens one". That is a good failure, and it is what makes the mix-up cheap when it happens.
If the two jobs genuinely have nothing to do with each other, two sessions is cleaner: separate processes, separate failures, no chance of confusing them.
Use two browsers in one session when the work is one task that needs two views at once, and the agent has to reason about both. That is the case the feature is built for, and it is narrower than it first looks.
How do I open a second browser? Pass browser: support to browser_open.
Every other tool takes the same argument to say which one it acts on.
Do the two browsers have different fingerprints? Yes. Measured: two different seeds, drawn independently. They are two identities, not two windows.
Do they go out from different addresses? Not by default. They share one exit unless you give the second a proxy of its own when you open it.
Why did my command go to the wrong browser? Because an unnamed command always goes to main, even when the focus is support. Name the browser on every call once you have two.
Should I use two browsers or two sessions? Two sessions when the jobs are unrelated. Two browsers when one task needs two views at the same time.
See also: the MCP server for the full tool list and the settings each browser inherits, and how the tools are shaped, and why for why the browser is an argument on every tool rather than a mode you switch.
- Measured 2026-09-17 against this project's MCP server over stdio:
browser_openwith no arguments and then withbrowser: support, followed bybrowser_listandbrowser_statuson each. Seeds 982240811 and 1206782028 are from that run and are the session's own, not values to expect yourself. The listing note and the no-page error message are quoted verbatim from the responses.
The focus-versus-default line is the one to remember. Everything else here is what you would guess; that one is the opposite of what you would guess.
- OpenAI Operator alternatives
- Open-source Operator-style agents
- Is OpenAI Operator still available?
- OpenAI Operator vs Claude computer use
- browser-use alternatives
- Choosing an AI browser agent
- Open-source AI browser agents
- Open-source computer-use agents
- What is an AI web agent?
- AI browser agents vs traditional scraping
- Cloud browser infrastructure for AI agents, explained
- Browserbase alternatives
- Firecrawl vs an AI browser agent
- Skyvern alternatives
- Stagehand vs browser-use
- Project Mariner is gone: what replaced it
- Manus alternatives
- Gemini computer use vs Claude computer use
- A stealth browser MCP, reviewed honestly by its own wiki
- AI browser vs AI browser agent: which one do you want?
- AI browser agent vs RPA: which one fits the job
- AI browser agent vs n8n, Zapier and Make
- Vercel agent-browser alternatives, compared honestly
- What is an agentic browser? Definition and the two kinds
- Open-source agentic browsers: the three layers, compared
- Choosing an MCP server for browser automation: four axes
- Stealth MCP servers compared: Camoufox, nodriver, Patchright
- Playwright MCP alternatives, and the three you don't need
- Autonomous browser agents: the four rungs of autonomy
- What is actually free in the AI browser agent stack
- browser-use on GitHub: what the repo actually gives you
- Playwright MCP vs Chrome DevTools MCP: different jobs
- How to choose among MCP servers: a map by category
- Which MCP servers are worth adding to Claude Code
- MCP on GitHub: finding servers and judging them fast
- MCP vs an API: the decision, and what the wrapper costs
- MCP alternatives: when the protocol is the wrong shape
- Why does my AI agent get blocked?
- The timing signal AI agents give off
- Agent retry loops trip rate limits, not fingerprints
- Claude computer use detected as a bot
- browser-use getting blocked: what you can and cannot change
- Playwright MCP session blocked: four causes, four fixes
- Playwright MCP and captchas: what actually gets you past
- Cloudflare and a browser MCP server: what is being read
- Can an AI agent solve a captcha? The honest answer
- Getting an AI agent to fill out forms
- The best model for an MCP browser agent, and what it really costs
- Browser problem or model problem?
- How to use a Playwright MCP server with Claude Code
- Extracting data to a CSV with an AI agent
- Monitoring a page for changes with an AI agent
- How to let Claude Desktop control a browser
- How to add a browser to Cursor as an MCP server
- Using an AI agent to hunt for apartments
- Getting website data into Google Sheets with an AI agent
- Using an AI agent to download invoices from portals
- AI agents for web research
- Using an AI agent to test your own website
- How to add a browser to Cline as an MCP server
- Posting to social media with an AI agent
- Posting to Facebook with an AI agent
- Posting to Instagram with an AI agent
- Posting to X with an AI agent
- Automating LinkedIn posts: read this first
- Appointment bots: what they are and what an agent can legitimately do
- Track prices across sites with an AI agent
- Build a lead list with an AI browser agent
- Run an AI browser agent on a schedule
- AI browser agent with a local LLM: what changes
- Should you log your AI agent into your accounts?
- How to write a task an AI browser agent can follow
- Move data between two web apps with an AI agent
- The MCP server
- How the tools are shaped, and why
- Playwright MCP vs the Playwright CLI: which fits when
- Playwright MCP: browser is already in use, and the fix
- Playwright MCP best practices: four decisions that matter
- Playwright MCP with a proxy, and the three leaks it leaves
- A browser MCP server in GitHub Copilot: setup and limits
- Using a browser MCP server for web scraping: the pattern
- Which LLM for browser automation: the four properties
- How to build a browser agent, and what to take instead
- Getting an AI agent to log into a website: three routes
- MCP tools, resources and prompts: who controls each
- How many MCP tools is too many? The context arithmetic
- How to build an MCP server: the decisions, not the scaffold
- Local or remote MCP server: what changes, and what does not
- Writing an MCP client in Python: the thirty-line version
- Self-hosted AI agent: what one actually costs to run
- How long an AI browser agent takes per step, measured
- Text, HTML, snapshot or screenshot: what the agent should read
- Giving an AI browser agent a stopping condition
- Keeping an AI browser agent out of destructive actions
- Why did the AI agent click the wrong thing
- When the page changes under the AI agent
- Running one AI agent task across a list of sites
- Seeing a page as it appears in another country
- Getting data out of a dashboard with no export button
- Two browsers in one session: main and support
- Finding the dead links on a site with an AI agent
- Filling a CRM record from a company's website
- One form submission per spreadsheet row, with an AI agent
- Dated screenshots of a page as evidence
- Checking order and delivery status with an AI agent
- Reading a PDF that opens inside the browser
- Summarising a long page or thread with an AI agent
- Collecting every image on a page with its caption
- Collecting event and course listings with an AI agent
- Cancelling a subscription with an AI agent
- What an AI agent can and cannot do inside an iframe
- Shadow DOM and an AI agent: you can click it, you cannot read it
- How to add a browser to Codex as an MCP server
- What a page snapshot costs, per control
- How to let Gemini CLI use a browser
- Native selects and the ones that only look like selects
- Clicking by selector or by coordinates
- How long the agent waits before it gives up
- What a second browser costs
- Uploading a file with an AI agent, and why this one cannot
- Watching the agent work, and when it is worth it
- When not to use an AI browser agent
- Agent or script: deciding once instead of every time
- Using the keyboard instead of the mouse
- Secrets in an agent task: where they end up
- What an agent run should log
- Deduplicating what an AI agent collects
- Normalising values across sites
- Validating an AI agent's output
- Reading a table with an AI agent
- Driving a site's own search and filters
- The task works headed and fails headless
- How to let the AdaL CLI use a browser