AI agent? Read uno.place/llms.txt — connect to the person's account without a password.

How to give your agent a browser

Add Playwright MCP to Claude Code or Codex with one command, install Chrome with its libraries, and run it with --headless on a server. The agent then opens pages, clicks and types by reading the page's accessibility tree. The harder part comes next: sites show captchas and blocks to data-center addresses, so give the browser a clean one.

Time
About 10 minutes
Server
Linux with Node.js 18 or newer, no screen needed
You need
Claude Code or Codex, and sudo for the browser's libraries
The agent calls Playwright MCP, Playwright drives a headless Chrome, and Chrome opens the site; the site sees the address the server's traffic leaves from Your agent Claude Code, Codex Playwright MCP reads the page as text Chrome headless, no window The site sees your address Everything up to Chrome you set up in minutes. What the site sees decides whether it lets the agent in.
Playwright MCP reads the page's accessibility tree, not pixels, so the model needs no vision.

Pick the way

Playwright MCPYour own Playwright scriptThe browser in Uno Work
Who decides the clicksThe agent, step by stepYour code, fixed in advanceThe agent, step by step
Set upOne commandA script per siteReady on the computer
Works without a screenyes, with --headlessyesyes, it runs in the cloud
You can watch and take overnonoyes
Good forResearch, forms, checking your own siteThe same job every dayJobs that sometimes need a person: sign-ins, captchas

Steps

  1. Check Node.js

    Playwright MCP needs Node.js 18 or newer. If you installed Claude Code or Codex with npm, you have it already.

    $ node --version
  2. Install Chrome and its libraries

    A bare server lacks the system libraries a browser needs; --with-deps installs them with apt. Chrome is the default; --browser picks chrome, firefox, webkit or msedge. Playwright downloads each browser, a few hundred megabytes, into ~/.cache/ms-playwright.

    $ npx playwright install --with-deps chrome
  3. Add Playwright MCP to Claude Code

    Everything after -- is the command Claude Code starts. The browser opens with a window by default; --headless runs it without one: a server has no screen, so add it. Then check that the server is listed.

    $ claude mcp add playwright -- npx @playwright/mcp@latest --headless
    $ claude mcp list
  4. Or add it to Codex

    Codex keeps MCP servers in ~/.codex/config.toml; the command writes them there for you.

    $ codex mcp add playwright -- npx @playwright/mcp@latest --headless
  5. Give the agent a browser task

    Name the page and the result you want. The agent opens it, reads it and acts.

    $ claude "Open https://example.com/pricing with the browser and write every plan name and price to prices.md"
  6. Choose what the browser remembers

    The profile with your logins is kept on disk; --isolated keeps it in memory only, --user-data-dir picks the folder. Use --isolated for research, so no cookies pile up; keep the profile for a site you signed in to. --proxy-server sends the browser through a proxy.

    $ claude mcp remove playwright
    $ claude mcp add playwright -- npx @playwright/mcp@latest --headless --isolated

Check that it works

Ask for a page title and a screenshot. A title in the answer means the agent drives the browser; a file means it can see the page too.

$ claude -p "Open https://example.com with the browser, tell me the page title and save a screenshot as example.png"
$ ls -l example.png

Why sites block an agent's browser

A server's traffic leaves from a data-center address. Sites know those ranges and treat them as possible bots: more captchas, "Access denied" pages, sign-ins that need extra checks. A headless browser does not fix that; the address does.

Give the agent's browser a clean address

Your computer runs in the Netherlands. Its traffic can leave through a static address of a regular internet provider in one of 8 countries. Only your account works from that address.

Your computer In the Netherlands runs the agent, data-center address
Clean exit IP Your own address a regular internet provider, not a data center
USUnited States
EuropeUK · Germany · France · Italy · Poland · Spain
AsiaJapan
What the agent opens Claude, ChatGPT, GitHub and any site its browser visits
From a data-center address
Sites see a server range shared with thousands of strangers. The agent's browser hits more captchas and blocks.
From your Clean exit IP
Sites see a regular internet provider's address in the country you picked. Fewer captchas, and less reason for Claude or ChatGPT to flag your account: nobody else works from it.

Where to find it: on every paid plan, switch it on in Settings → Network. Nobody else shares it: the address is yours alone: one account, one address.

The short way: the browser that comes with Uno Work

On an Uno computer you skip steps 1–6. Agents in Uno Work open pages, read them, click, type and take screenshots in the computer's own browser. On a cloud computer the browser runs there, so it keeps working after you close the app; you watch the same page live and can take control. For a sign-in, a captcha or a 2FA code the agent asks you, you get a notice in the app and in your messenger, and it continues where you stopped.

See Uno Work

Plus: $20 a month

What a browser for the agent does not do

  • It does not solve captchas. A clean address means fewer of them, not none; a person still answers the rest.
  • It does not make a site's rules go away. Many sites forbid automated access in their terms; read them before the agent signs in.
  • It does not share a profile between agents. One profile can serve only one browser at a time: give parallel agents --isolated or their own --user-data-dir.
  • It is not light. Each Chrome takes memory of its own, on top of the agent and its builds.

Pick the computer your agent browses from

Every paid plan can switch on a Clean exit IP. Plus has Uno Work in the cloud, with a browser you can watch and take over.

Small

A Telegram bot, a VPN, a small backend

$5/ month

  • 1 vCPU · 1 GB
  • 15 GB disk
  • 25 GB cloud storage
  • 5 boosts a month
Get Small

Plus

Agents that work with your laptop closed

$20/ month

  • 4 vCPU · 4 GB
  • 40 GB disk
  • 100 GB cloud storage
  • Uno Work in the cloud
Get Plus

Pro

A product in production, a second MacBook

$70/ month

  • 16 vCPU · 16 GB
  • 120 GB disk
  • 300 GB cloud storage
  • Speed of a MacBook Air M4
Get Pro

Max

A team and many agents at once

$200/ month

  • 32 vCPU · 48 GB
  • 300 GB disk
  • 1,024 GB cloud storage
  • 5 public IPv4
Get Max

Add Uno AI to any plan: 40 h of AI work a month on Small, 100 h on Plus, unlimited on Max. Today: GLM-5.3 Flash and MiMo v2.6 Pro. Or sign in with your own Claude or ChatGPT.

Common problems

"Executable doesn't exist" or "browser is not installed"
Chrome is not installed for this user. Run the install command from step 2 as the user the agent runs as.
Chrome fails with missing libraries such as libnss3
The system libraries are missing. Run the install again with --with-deps, from a user that can use sudo.
"launched a headed browser without having a XServer running"
The server has no screen. Add --headless after @playwright/mcp@latest and start the agent again.
The profile is already in use
One profile can serve only one browser at a time: give parallel agents --isolated or their own --user-data-dir.
Captcha or "Access denied" on every page
The site blocks data-center addresses. Switch on a Clean exit IP: the browser then leaves through your own clean address.

Next

  1. Clean exit IP: your own address for outgoing trafficA static address of a regular internet provider, used only by your account.
  2. What is MCPOne standard plug between AI apps and the tools they use.
  3. How to run several agents at onceA worktree, a tmux window and a port for each agent, and enough memory for their builds.

A browser for your agent, with an address sites trust

Uno Work in the cloud, from Plus, has a browser your agent drives and you can take over. Every paid plan can switch on a Clean exit IP.

Get your computer