AI that acts on its own is already on your devices

AI Agent Blocklist: block AI browsers, AI agents and agent extensions

AI browsers, agents, agent extensions, coding agents and MCP servers reach company laptops faster than IT can review them. They read mail, click buttons and run commands with your users' access. One daily blocklist puts them under your control again.

  sample entries, one daily feed
domainmanus.imGraded
extensioncecngibhkljoiafhjfmcgbmikfogdikoGraded
mac appai.perplexity.cometGraded
win appBrowserOS.BrowserOSGraded
cli@google/gemini-cliGraded
domainbrowser-use.comGraded
Next step

The AI Agent Blocklist: 2,600+ agent domains and 5,300+ device identifiers, updated every day

Ready files for firewalls, DNS filters, secure web gateways, Intune, Group Policy, the Google Admin console and Jamf.

From $99 a month for one organisation

How to stop AI agents that arrive faster than you can review them

A new AI browser or agent extension takes a user about two minutes to install. A security review takes weeks. In between, the agent works with real accounts.

Install

A user downloads an AI browser or adds an agent extension. Many install per user, without admin rights.

Sign in

The agent inherits the user's logged-in sessions: mail, CRM, file shares, banking.

Connect

The user links mail, calendar or drives, or starts an MCP server that opens a system to the agent.

Act

The agent reads pages, fills forms, sends messages and runs commands for the user.

Unseen

Nothing of this passed a review, a data agreement or your logging.

AI browsers

Full browsers with an assistant that reads every tab and an agent that acts in logged-in sessions.

AI agents

Autonomous agents and personal assistants that work through many steps in the user's accounts.

Agent extensions

Chrome and Edge extensions that bring an agent into the browser you already manage.

Coding agents

Command line and IDE agents with access to source code, secrets and a shell.

MCP servers

Connectors that hand an agent the keys to databases, mailboxes and cloud consoles.

Agent platforms

Web services where staff build agents that run on a schedule with company accounts.

Gartner told security leaders to block AI browsers for now. Its December 2025 advisory "Cybersecurity Must Block AI Browsers for Now" advised blocking AI browser installs with existing network and endpoint controls. Read what Gartner's advice means in practice.

What is at stake when agents run unmanaged

The risk is not the AI vendor. It is AI that acts with your users' access, outside your controls, on pages and documents anyone can write.

Data leaves

Pages, tabs, mail and files go to an AI service to be processed, often under consumer terms.

Steered agents

Researchers showed in 2025 that hidden text on a web page can steer AI browser assistants into sending out data from other tabs.

Wrong actions

An agent misreads a page and submits a wrong order, form or payment in a real account.

Destructive commands

Coding agents with shell access can delete files or change systems after one bad instruction.

Signs agents are already in use

  • DNS or proxy logs show agent services nobody approved
  • Browser reports list new extensions with access to all sites
  • Software inventory shows browsers you never packaged
  • Developer laptops run AI command line tools from npm or PyPI
  • AI notetakers appear in meeting invites
  • Tokens and app consents exist for tools with no owner

For a structured first look, follow our guide to finding shadow AI agents.

Guides for every way AI agents get in

Each part of this site starts from one problem and ends with the files and settings that fix it. Pick the door that worries you most.

AI browsers

Comet, Dia, Opera Neon, Genspark and others, plus the AI now built into Chrome and Edge. How to block them, remove them and answer "is it safe".

AI agents

Autonomous agents, personal assistants, agent modes in chat apps and AI bots in meetings. How to build an inventory, set governance and block what you did not approve.

Agent extensions

Extensions that read and change every page. Block and allow lists for Chrome, Edge and Firefox in every major console.

Coding agents

Claude Code, Codex CLI, Cursor, GitHub Copilot and the CLIs around them. Control what runs and protect secrets.

MCP servers

The connectors that give agents real access to data and systems. Risks, block lists, allowlists and a checklist for approving the few you need.

Firewalls and DNS

Load the list into FortiGate, Palo Alto Networks, Zscaler, Cisco Umbrella, Sophos, WatchGuard, UniFi, Pi-hole or AdGuard Home.

Intune and Chrome policies

Device and browser policies that work off your network: Intune, Group Policy, the Google Admin console and Microsoft Defender.

Guides

Acceptable use policy, risk assessment, allowlist or blocklist, and pages for schools with Chromebooks.

What is in the AI Agent Blocklist

One dataset, built from our AI tools data of more than 12,000 AI tool domains, narrowed to AI that acts on its own and the tools that build it.

2,600+

AI agent domains

Every domain carries its tool name and categories:

AI browsers and browser agentsGeneral autonomous agentsPersonal AI assistantsAgent building platformsMulti-agent orchestrationMCP serversAutonomous coding agentsDeep research agentsMeeting and voice agentsAI sales agentsAI RPA
5,300+

Agent identifiers on devices

The IDs your browser and device policies need:

Chrome extension IDsEdge add-on IDsFirefox add-on IDsmacOS bundle IDsWindows package IDsProcess namesCoding CLI package names

Every entry has a risk grade:

LowModerateElevatedHighSevere

Every entry is checked before it enters the list, and new AI browsers, agents, extensions and MCP servers are added as they appear. A tool that sits in several categories lists all of them, so you can filter the way your policy is written.

Columns, real sample rows and every format are on the AI Agent Blocklist page.

How it plugs into your controls

No agent to install and no traffic sent to us. You load files into the tools you already run, and the daily updates keep them current.

Network

Add the domain list as an external list or threat feed in your firewall, or as a custom block list in your DNS filter.

Browser

Paste the extension IDs into the extension blocklist policy for Chrome and Edge through Group Policy, Intune or the Google Admin console.

Device

Use bundle IDs, package IDs and process names in Jamf Pro, Intune or your other device management to block and remove AI browsers and agents.

Review

Use the CSV or JSON as a lookup in your logs and inventory. See who uses which agent before you widen the block.

FortiGatePalo Alto NetworksZscaler Internet AccessCisco UmbrellaSophos FirewallWatchGuardUniFiPi-holeAdGuard HomeMicrosoft IntuneGroup PolicyGoogle Admin consoleJamf ProKandjiMosyle

Block high risk AI agents first, then decide on the rest

Most organisations do not block every AI tool. They block AI that acts, keep the assistants they licensed, and watch the grey zone.

Block now

  • AI browsers on managed devices
  • Agent extensions with access to all sites
  • Unapproved coding CLIs and public MCP hubs
  • Every High and Severe entry

Monitor

  • Moderate and Low agents in categories you have not ruled on
  • Meeting notetakers and voice agents
  • Agent platforms used by single teams

Allow

  • Assistants and coding tools you licensed
  • Their enterprise settings switched on
  • Removed from your copy of the block files

Who uses an AI agent blocklist

The same files serve different jobs. Here is what each role usually needs from them in the first month.

IT and security admins

Files that load without conversion, so the block is live the same day. Extension IDs and app identifiers they would otherwise look up one by one.

CISOs

A clear line on AI that acts: what is blocked, what is allowed, and a risk grade per tool to explain the decision to the board and to auditors.

MSPs

Separate files per client, so a law firm and a design studio get different block levels from one source that updates itself.

Compliance owners

Proof that unapproved AI agents cannot reach regulated data, plus a CSV that maps every blocked tool to its category.

School and district IT

AI browsers and agent extensions kept off student Chromebooks and staff laptops, before an agent completes homework or tests for a student.

Product makers

AI agent data inside a firewall, filter, browser or MDM product, updated daily, under a product licence.

Your first week with the blocklist

A calm order that blocks the riskiest tools first and gives teams time to ask for the AI they really need.

Day 1

Load the CSV into your log search. Find which agents, AI browsers and extensions are already in use, and by which teams.

Day 2

Add the High and Severe domains to the firewall or DNS filter. Tell staff which approved AI tools stay available.

Day 3

Push the extension blocklist to Chrome and Edge. Agent extensions stop installing and installed ones are switched off.

Day 4

Use app identifiers and process names in your MDM to block AI browsers and desktop agents, and remove them where policy says so.

Day 5

Review requests, extend the block to Elevated entries if your policy allows, and schedule the weekly policy refresh.

Need the policy text that goes with it? Start from our AI agent acceptable use policy guide.

AI agents inside the browsers you already allow

Blocking new AI browsers is half the job. Chrome and Edge now ship assistant and agent features of their own, and agent extensions add more. These are controlled by policy, not by a domain list.

WhereWhat it isHow you control it
Gemini in ChromeGoogle's assistant in Chrome, with an agent mode that can act on web pagesChrome policies GeminiSettings and GeminiActOnWebSettings. Check the current Chrome Enterprise policy list.
Copilot in EdgeMicrosoft's assistant in Edge, including Copilot ModeEdge policies such as EdgeCopilotEnabled. Check the current Edge policy list for Copilot Mode.
Claude in ChromeAn extension that lets an AI agent browse and act in ChromeExtension ID fcoeoabgfenejglbffodgkkbkcdhcgfn in your extension blocklist or allowlist
Other agent extensionsHundreds of extensions that add agents to Chrome, Edge and FirefoxThe extension block lists in the AI Agent Blocklist

Why a domain list alone does not stop agents

Network blocks are the fastest first step. But agents also live in places a network rule never sees.

Off the network

A laptop at home or on hotel Wi-Fi skips your firewall. Browser and device policies still apply there.

Inside the browser

An agent extension can use the same sites your staff need. Blocking its extension ID removes it without blocking those sites.

Installed apps

AI browsers and desktop agents are installed software. Bundle IDs, package IDs and process names let MDM find and stop them.

Developer tools

Coding agents arrive as packages. Package names support rules in package proxies and checks on developer machines.

Rather not manage files? We run the AI Agent Filter for you

The hosted AI Agent Filter is a DNS filter with one switch that blocks AI agents and AI browsers. You point your DNS at it, and we keep the 2,600+ agent domains current every day.

Point your DNS

Set our two DNS addresses on the router or firewall. Every device on the network is covered, with nothing to install.

Turn on the switch

"Block AI agents and AI browsers" in a profile. Tools you approve by name stay open.

See the reports

Every AI tool and agent your networks reached, with its type and the rule that decided it.

Hosted filter

AI Agent Filter

The AI agents switch, AI tools controls, 58 web categories, a dashboard and reports.

From $99 a month for one organisation

How the filter works See the plan

The filter blocks the websites and services agents depend on. Apps and extensions already on devices need device policies, which is what the identifiers in the AI Agent Blocklist are for.

Need every web category, not only AI agents?

Teams that run their own firewall or DNS filter often want the rest of their web policy from the same source. The 20M filtering domains database adds 58 categories to the agent list.

AI Agent Blocklist + 20M Filtering Domains Database

The 20 million most visited domains, each with one or more of 58 filtering categories, in one CSV file updated daily or monthly. Quote on request for one organisation.

Plans for organisations, MSPs and product makers

Every plan includes all domains, all identifiers, every format and daily updates. Full prices are on the pricing page.

One organisation

One company, school or district. From $99 a month for one organisation.

Organisation plan

MSPs and multi-site

Separate files per client or site, for providers and organisations with many sites.

MSP plan

Product licence

Embed the data in a security, browser, filter or MDM product.

Product licence

One list for every door AI agents use

Domains for the network, IDs for the browser, identifiers for the device. Checked before they enter the list and updated every day.

  • Files for firewalls, DNS filters, MDM and browser policies
  • Categories and risk grades on every entry
  • Download access set up within one business day of your order

Ask us or order a plan

Tell us what you need to block and where your rules live. We reply within one business day.

  • Files for firewalls, DNS filters, secure web gateways, MDM and browser policies
  • Updated every day as new AI browsers, agents and extensions appear
  • Plans for one organisation, for MSPs and for product makers, or a hosted filter we run for you

We use your details only to answer you. No newsletters.

Questions about blocking AI agents

What is an AI blocklist?

An AI blocklist is a list of AI tools that your security controls refuse to reach or run. The AI Agent Blocklist focuses on AI that acts on its own: AI browsers, agents, agent extensions, coding agents and MCP servers. It holds both their domains and the identifiers they leave on devices.

How do I block AI agents?

Block agent domains at the firewall or DNS filter, block agent extensions with a browser extension blocklist, and block agent apps and command line tools with your device management. Keep a short allowlist of the AI tools you approved.

How do I stop AI agents without stopping all AI?

Use the categories and risk grades. Block AI browsers, agent extensions and High or Severe tools first, keep your licensed assistants on an allowlist, and review the rest with the teams that use them.

Is an AI domain blocklist enough?

No. A domain list only works while a device uses your network or DNS. Extensions and installed apps also need policy blocks on the device. That is why the AI Agent Blocklist ships extension, app and package identifiers next to the domains.

Can I block only high risk AI agents?

Yes. Each entry has a risk grade from Low to Severe. Load only High and Severe entries at first, or everything from Elevated upwards.

Does the AI Agent Blocklist cover generative AI tools in general?

It covers AI agents and the tools around them, built from our AI tools data of more than 12,000 AI tool domains. If you need a wider gen AI block, ask us through the form.

Does the blocklist remove AI agents that are already installed?

The files tell your tools what to block. Chrome and Edge switch off blocklisted extensions that are already installed. Installed AI browsers and agent apps need an action in your device management, such as Jamf Restricted Software or an Intune uninstall script.

How much does it cost?

Plans start from $99 a month for one organisation. MSPs and product makers have their own plans. All prices are on the pricing page.

Perplexity and Comet are trademarks of Perplexity AI, Inc. Manus is a trademark of its owner. Dia, Opera Neon, Genspark, BrowserOS and Browser Use are trademarks of their respective owners. Claude and Claude Code are trademarks of Anthropic, PBC. Codex is a trademark of OpenAI. Cursor is a trademark of Anysphere, Inc. GitHub Copilot is a trademark of GitHub, Inc. Gemini, Chrome and Google are trademarks of Google LLC.

Microsoft, Edge, Intune and Defender are trademarks of Microsoft Corporation. Firefox is a trademark of the Mozilla Foundation. FortiGate is a trademark of Fortinet, Inc. Palo Alto Networks is a trademark of Palo Alto Networks, Inc. Zscaler is a trademark of Zscaler, Inc. Cisco Umbrella is a trademark of Cisco Systems, Inc. Sophos, WatchGuard, UniFi, Pi-hole, AdGuard, Jamf, Kandji and Mosyle are trademarks of their respective owners. Gartner is a registered trademark of Gartner, Inc.

Names are used only to identify the products discussed. AI Agent Blocklist is not affiliated with, sponsored or endorsed by these companies.