Repository navigation
what is an agentic browser
An agentic browser is a browser that can carry out a task for you, rather than only display pages while you do the clicking. You give it an instruction in words, and it navigates, reads, clicks, fills forms and reports back. The browser is still a browser: what is new is that an AI agent sits inside it with the ability to drive the same UI you would have driven.
That is the whole definition. Everything else on this page is the part the definition hides, and the reason people search this term twice.
Search this phrase and you get two kinds of result that barely overlap: consumer browser reviews, and security advisories. They are not talking past each other by accident. There really are two things.
A browser you use, with an agent inside it. You install it, you log in, you browse. When you want something done, you ask. Perplexity's Comet, OpenAI's ChatGPT Atlas, Opera Neon, Fellou, Dia, Sigma and the open-source BrowserOS all sit here, as do the agent features being folded into Chrome and Firefox. The agent inherits your session, your cookies and your logged-in accounts, because it is running in your browser.
A browser your code drives, with an agent in the loop. No window you sit in front of. A program starts a browser, an LLM decides what to do next, and the browser does it. browser-use, Skyvern, Stagehand and invisible_playwright_mcp - this wiki's project - are this shape. Nobody is logged in unless you arrange it.
The words are almost the same and the risk profile is opposite. If you are comparing tools, work out which one you meant first, because a review of Comet tells you nothing useful about a library, and vice versa. This wiki keeps that split on its own page: AI browser vs AI browser agent.
Whichever shape it is, the same four problems show up, and how a product answers them is a better comparison axis than its feature list.
How the agent perceives the page. Two families. Either the model reads a structured description of the page - the accessibility tree, or the DOM reduced to something a model can hold - or it looks at a screenshot and works in pixels. Structured reading is cheaper and more reliable when the page is well built; pixels are the fallback for canvas, custom widgets and anything the tree does not describe. Most serious tools now do both and pick per step.
How it acts. A selector-based click is precise and brittle. A coordinate click is general and blind. Whether the resulting event looks like a person depends on the layer underneath, which is a separate question from the agent.
What it is allowed to do. This is the one that will decide the category. An agent that inherits your logged-in session can do anything you can do, including things you did not ask for, if it is talked into it by text on a page.
Whether the site can tell. An agentic browser is still automation as far as a defended site is concerned, and the tells live in the browser build and the network path, not in the agent's reasoning. This wiki has a whole section on that: when the agent gets blocked.
The reason a security vendor outranks a product page for this term is that the consumer shape has a genuinely unsolved problem: prompt injection. The agent reads the page in order to act on it, so text on the page arrives in the same channel as your instruction. A page can contain a sentence addressed to the agent rather than to you. If the agent is browsing with your session, the worst case is not a wrong answer, it is an action taken with your credentials.
The research literature in 2026 is actively proposing browser-level answers to this - work on a same-origin policy for agentic browsers is one direction, the idea being that an agent's authority should be scoped the way a script's already is. None of it is settled. Treat any product claim of "we handle prompt injection" as a claim, and ask what it means mechanically.
The library shape dodges most of this by default, for a boring reason: it usually starts with no session at all. That is not virtue, it is scope. The moment you hand a code-driven agent a logged-in profile, you have the same problem, which is why this wiki has a page asking whether you should log your agent into accounts before you do it.
- You want to save yourself clicks on sites you are logged into. A consumer agentic browser. Accept that it acts with your session, and read what it does with your data before you enable it on your bank.
- You want a task to run repeatedly, unattended, on a schedule. Not a consumer browser. You want the library shape, where the run is a program you can log, replay and rate-limit.
- You want to pull data out of sites at volume. Probably not an agent at all, at least not for the whole job. See AI browser agents vs traditional scraping: an agent is the right tool for the part that changes and the wrong tool for the part that repeats.
- You want your existing assistant to be able to open a page. That is the MCP route, and it is the cheapest of the four to try: the MCP server.
What does agentic mean in a browser? That the browser can take a sequence of actions toward a goal you stated once, instead of executing the single action you just clicked.
Is an agentic browser the same as an AI browser? Not quite. An AI browser may only summarise and answer; an agentic one acts. Products blur the line because acting is the feature people pay for.
Is an agentic browser safe? The mechanism to worry about is prompt injection, not the AI generally. The exposure scales with what the agent is logged into, so the honest answer depends entirely on how you configure it.
Are there open-source agentic browsers? Yes, in both shapes. Open-source agentic browsers and agents covers what exists and what each one actually gives you.
Will sites block an agentic browser? Defended sites block automation whether or not there is an agent involved, and they judge the browser and the network path, not the intent. Start with why an agent gets blocked.
See also: Choosing an AI browser agent for the decision framework, What is an AI web agent? for the library shape in detail, and the MCP server for turning an assistant you already run into one.
- Top 5 agentic browsers in 2026: capabilities and security risks, Seraphic Security, retrieved 2026-09-10, for the security framing and the product list.
- What is an agentic browser, Sigma, retrieved 2026-09-10, a vendor overview of the consumer shape.
- Same-Origin Policy for Agentic Browsers, arXiv preprint, retrieved 2026-09-10, for the browser-level direction on prompt injection.
- The browser-use, Skyvern and invisible_playwright_mcp repositories for the library shape.
Written while maintaining invisible_playwright_mcp, which is one of the tools in the second category. The page is deliberately about the category rather than the product, and the section that tells you when you do not want an agent is there because it is the honest answer more often than a project in this space likes to admit.
- 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