Verdict for IT teams

Opera Neon Security: Should You Block It at Work?

Opera Neon is an AI browser whose agents act inside the user's signed in sessions. Since August 2026 outside AI agents can also connect to it on the free plan. Decide before it spreads across your fleet.

Managed fleet verdictBlock by default
  • Agents act in logged in sessionsYes
  • Outside agents can connectYes, MCP and CLI
  • Admin console or policiesNone published
  • Built in VPNYes, free
  • Passwords and cookies stay localOpera says yes
  • Training on chats and historyOpera says no

Next step

AI Agent Blocklist

Block AI browsers like Opera Neon and the agents that connect to them, with daily domain lists for firewalls and DNS filters and identifier files for Intune, Group Policy and Jamf.

From $99 a month for one organisation

What Neon is now

Opera Neon in one minute

Neon is not a sidebar added to Opera's regular browser. It is a separate desktop browser built around agents, with its own plans and its own way for outside AI tools to plug in.

30 Sep 2025

Opera ships Neon to its waitlist. It opens to everyone without a waitlist in December 2025.

Windows, macOS

Opera lists Neon for Windows and Mac computers only. It is a desktop browser.

$19.90 a month

The Standard plan, billed month to month, unlocks Neon's own agents.

Free since Aug 2026

A free plan includes the browser and the connections for outside agents.

Free plan

$0
  • The Neon browser for Windows or macOS
  • Built in ad blocker
  • Free VPN with no login
  • MCP Connector and Browser CLI for outside agents
  • Basic AI chat to summarise pages and compare products

The free plan is what makes Neon an IT question in every team, not only for people who pay.

Standard plan

$19.90 / month
  • Neon AI agents that act on the web
  • Deep research
  • Image and video generation, code with deploy options
  • Skill cards and scheduled tasks
  • Long memory and Google connectors

Prices and features from Opera's subscription page. Check it before you write policy, as plans change.

The problem in detail

Is Opera Neon safe? Capability against risk

Opera makes privacy promises for the individual user. A company asks a different question: what can this browser do with work accounts, and what can IT see or switch off. This matrix answers that, feature by feature.

CapabilityWhat it reachesRisk for a companyCentral controlWhat to do
Neon DoTabs in the user's own session, where they are already signed inClicks and form entries made as the user inside work appsNone publishedNeeds a control
TasksA separate workspace of tabs for one jobPage text from many sources feeds one agent runNone publishedReview
Cards and skill cardsReusable instructions, including a community storeInstructions written by strangers run on work pagesNone publishedReview
Cloud buildsA cloud workspace that Opera said at launch runs on its European serversWork content leaves your tenancy; results can be shared as a linkNetwork rules onlyReview
MCP ConnectorOpen tabs, page content and signed in sessionsAny connected outside agent acts inside the user's sessionsNone publishedNeeds a control
Browser CLILocal control of the browser, no account neededScripts and coding agents drive a signed in browserEndpoint rules onlyNeeds a control
Built in VPNBrowser trafficDomain rules on DNS and firewall may not see the sites visitedEndpoint rules onlyNeeds a control
Scheduled tasks and long memoryRepeated runs and stored contextAgent work happens when nobody is watching the screenNone publishedReview
Credentials and cookiesThe deviceOpera says sensitive data stays local, which narrows one riskNot neededLittle action
Needs a control acts as the user or routes traffic past network rulesReview moves content or instructionsLittle action vendor control in place

Who is exposed first

Developers

They run coding agents that can call a local CLI. A browser signed in to cloud consoles is the prize.

Marketing and sales

Video, image and site builds appeal to them, and so does paying $19.90 on a card.

IT and admins

Their sessions reach admin consoles. An agent acting in that browser acts with admin rights.

Finance and HR

Their tabs hold payroll, invoices and personal data that should never feed an outside agent.

What Opera promises, and what that leaves open for a company

Opera's statements are real limits on what Neon does with data. Read them next to the questions an auditor would ask about a browser on a work laptop. White cells show what Opera says; amber cells show what that leaves open.

Opera says
What it leaves open
Passwords, card details and cookies stay on the device.Nothing sensitive is copied to a cloud agent to log in for you.
The agent still uses those sessions.Work that Neon Do or a connected agent does is done as the signed in user, with that user's rights.
Chats, history and account data are not used for training.Opera also says they are not used for personalisation or advertising.
Prompts still reach cloud models.Neon's plans list models from several AI providers. What an agent reads on a page has to be sent to a model to be processed.
Cloud builds run on Opera's European servers.Opera gave that location for its cloud workspace at launch.
No admin view of that workspace.Opera publishes no export, retention or deletion controls for an organisation, because there is no organisation account.
Only authorised clients reach the MCP Connector.Access runs through a private server address.
The user does the authorising.Which assistant or automation tool gets a live, signed in browser is the employee's choice, not yours.

Researchers showed in 2025 that hidden text on a web page can steer the agent in an AI browser. Treat this as a property of every agentic browser, not of one product, and plan your controls around it. See how prompt injection works in AI browsers.

The part most teams miss

Opera Neon MCP and CLI: the browser as a tool for other agents

Neon's own agents are only half the story. Opera added an MCP Connector in March 2026 and made it and a local Browser CLI free in August 2026. Any agent the user already runs can now borrow a live, signed in browser.

Outside agent

A chat assistant, an automation tool or a coding agent on the laptop

Neon connection

MCP with an Opera account, or the local CLI with no account

Work sessions

Mail, CRM, cloud consoles and anything else open in Neon

MCP Connector Opera account

  • Outside AI clients read open tabs and page content
  • They work inside signed in sessions
  • They navigate, fill forms, open tabs, search and take screenshots
  • Access runs through a private server address given to the client

Browser CLI No account

  • Connects on the same computer
  • Built so agents use fewer tokens when they browse
  • Fits coding agents and scripts that already run in a terminal
  • Needs no Opera account, so there is no sign-in to watch for

MCP risks in general, from tool poisoning to over-broad access, are covered on MCP security risks. Agents that drive a screen or browser are covered on computer use agents.

Opera said in March 2026 that a simpler browser connector would also come to Opera One and Opera GX. If you allow those browsers, check whether it has shipped and what it can reach.

How we solve it

Three options: block, monitor or allow with conditions

Opera lists no business plan, admin console or policy set for Neon. That leaves your own endpoint and network rules as the controls. Gartner's December 2025 advice, "Cybersecurity Must Block AI Browsers for Now", points the same way for most organisations.

Option 1, defaultBlock

For every managed device unless a team has a named owner and a reason.

  1. Endpoint rule so Neon cannot run
  2. DNS and firewall rule for operaneon.com and the AI browser list
  3. Block the MCP and CLI connections in the AI clients you manage
  4. Inventory alert if Neon appears again
Option 2, short termMonitor

For the first weeks while you find out who uses it and why.

  1. Endpoint rules in audit mode
  2. DNS logging for operaneon.com
  3. App inventory report each week
  4. Set an end date, then move to block
Option 3, rareAllow with conditions

For research or product teams that must test agentic browsers.

  1. A separate device or virtual machine
  2. No SSO sessions to production systems
  3. No MCP or CLI connections from work agents
  4. Written approval in your AI agent use policy

Windows with Group Policy: AppLocker

  1. Open the Group Policy Management Console and edit the GPO for your computers.
  2. Under Computer Configuration, Windows Settings, Security Settings, Application Control Policies, AppLocker, select the executable rule collection.
  3. Choose Create New Rule, set Permissions to Deny and pick the Publisher condition.
  4. Browse to a signed Neon file on a test PC and move the slider to cover the product, not one version.
  5. Under System Services, set Application Identity to start automatically. AppLocker is not enforced without it.
  6. Run in audit only mode first, then enforce.

Console labels differ between Windows versions, so check the current console. On Intune, the same block goes into an App Control for Business policy: see block AI with Intune.

Macs with Jamf Pro

  1. Open Computers, Restricted Software, New.
  2. Name the record, for example "AI browser: Opera Neon".
  3. In Process Name, enter the exact app name. To read it, open Neon on a test Mac and run ps -acx.
  4. Tick Kill Process. Add Send email notification on violation for the first weeks.
  5. Scope it to all managed Macs and save. It applies at the next check-in.

Labels in Jamf Pro change between releases. Check the current admin console before you share a runbook.

Firewall and DNS

  1. Block operaneon.com, Neon's main domain, so its site and download page fail.
  2. Add the AI Agent Blocklist as an external or custom list so other AI browsers fail too.
  3. Block or inspect browser VPN traffic in line with your network policy.
  4. Review blocked lookups weekly to find devices that still try.

Vendor by vendor steps: block AI agents at the firewall.

Signs Neon is already in use

  1. Opera Neon in Windows or Mac app inventory
  2. DNS lookups for operaneon.com from staff devices
  3. Expense claims for a monthly AI browser subscription
  4. Agent tools configured with a browser MCP server address
  5. Browser traffic that leaves through a VPN you did not deploy

If you find it, remove it with your MDM: remove AI browsers from managed devices.

A one week rollout that does not break anyone's day

  1. Day 1Pull app inventory from your MDM and DNS logs for operaneon.com. Note who uses Neon and on which devices.
  2. Days 2 to 3Deploy the endpoint rules in audit mode. Tell the users you found what changes and why, with a named alternative.
  3. Days 4 to 5Switch the rules to enforce, add the domain list to DNS and firewall, and block MCP and CLI connections in managed agent tools.
  4. Day 7 onwardsReview blocked attempts weekly. Handle exception requests through your written AI agent policy, not by ticket.

What the list adds

Block Neon and the AI browsers people try next

A rule for one browser closes one door. The AI Agent Blocklist covers AI browsers, browser agents, agent extensions and MCP servers, so the next download fails the same way. Every entry carries a risk grade, so you can start with the higher grades.

AI browser domains for DNS filters and firewalls
Windows package IDs and process names for AppLocker and Intune
macOS bundle IDs and process names for Jamf
Agent extension IDs for Chrome, Edge and Firefox policies
MCP server and agent platform domains
New entries every day as AI browsers appear

Microsoft Store ID

XPFFVLPJRVQTKM

Comet Browser on Windows, risk grade Moderate

Homebrew cask

browseros

BrowserOS on macOS, risk grade Moderate

Chrome extension ID

hmdkfngkncmbgiehonfebecighedjiak

Genspark extension, risk grade Moderate

Three real entries from the list. The full set of AI browsers we track is on the AI browser list, and the wider risks are summed up on AI browser security risks.

The product

AI Agent Blocklist: one download for every place you enforce

2,600+ AI agent domains and 5,300+ agent identifiers on devices, each with the tool name, category, identifier type and risk grade. Every entry is checked before it enters the list, and the lists update every day.

Plain domain list and hosts file
PAC file
Firewall external list format
DNS filter import files
CSV and JSON with all columns
Browser and device policy files for Intune, Group Policy, Jamf and the Google Admin console

After an order we set up your download access within one business day. Plans start from $99 a month for one organisation; MSP and product licences are on the pricing page.

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.

FAQ

Opera Neon questions from IT teams

Is Opera Neon safe for work?

Opera says credentials, card details and cookies stay on the device and that it does not train on chats or history. The work risk is different: its agents act in signed in sessions, outside agents can connect, and there is no admin console. On managed fleets we block it by default.

Is Opera Neon free?

Since August 2026 there is a free plan with the browser, ad blocker, VPN, basic AI chat and the MCP Connector and Browser CLI for outside agents. The Standard plan at $19.90 a month adds Neon's own agents, deep research, skill cards, scheduled tasks and long memory.

What is the Opera Neon MCP connector?

It lets outside AI clients work inside the live browser session: read open tabs and pages, use signed in sessions, navigate, fill forms, take screenshots and open tabs. It launched for subscribers in March 2026, became free in August 2026 and needs an Opera account.

What is the Opera Neon CLI?

The Opera Browser CLI is a local connection that lets agents and scripts on the same computer operate the browser. Opera says it needs no account. On developer machines this means a coding agent can drive a browser that is signed in to work systems.

Does the Opera Neon VPN get around our web filter?

It can. Neon has a free built in VPN with no login. Traffic sent through a VPN reaches your network as traffic to the VPN service, so DNS and firewall rules that read site names may not see where the browser goes. Block the app on managed devices so the question does not arise.

Is there Opera Neon for iPad, iOS or Android?

Opera lists Neon for Windows and macOS. Treat any phone or tablet app or APK file using the Neon name as unofficial until you have checked the publisher, and block installer files that do not come from Opera.

Is this the old Opera Neon from 2017?

No. Opera used the Neon name for a concept browser in January 2017. The current Opera Neon is a different, agentic AI browser that Opera announced in May 2025 and shipped on 30 September 2025.

Opera, Opera Neon, Opera One and Opera GX are trademarks of Opera Norway AS. Perplexity and Comet are trademarks of Perplexity AI, Inc. Genspark and BrowserOS are trademarks of their owners.

Microsoft, Windows, Intune and AppLocker are trademarks of Microsoft Corporation. Mac and macOS are trademarks of Apple Inc. Google is a trademark of Google LLC. Jamf and Jamf Pro are trademarks of Jamf. Gartner is a 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.