Opera brings AI context and agents into the browser

Opera makes the browser itself an AI workspace. Its mainstream Opera AI features help with questions and page context, while the separate Opera Neon browser adds agent-oriented research and actions. The distinction matters: summarizing an open article, connecting an existing agent and delegating a task to Neon’s built-in agents have different access requirements and different consequences for the user.

01 / ProductOne company, several levels of browser assistance

Opera is an active browser company, with Opera Limited identified on its investor site. Its current browser AI page presents Opera AI as free assistance that can use page and tab-group context. It also distinguishes access to third-party assistants in the sidebar. A convenient sidebar entry does not imply that another provider’s paid subscription is included.

The separate Opera Neon site presents a browser built around research, creation and task execution. Its current commercial page distinguishes a free browser tier from Standard, which unlocks built-in agent capabilities. Keep these products separate during evaluation: evidence about a feature in Neon should not be generalized to every Opera browser.

The AI relationship is substantial rather than cosmetic. Both products work with browsing context, and Neon can interact with web pages. That creates a useful opportunity to reduce copying between tabs and assistants, but it also makes the scope of visible context and authorized actions central to the decision.

02 / AudienceUseful when the work already lives across tabs

A plausible user compares public sources, gathers information from several web applications or repeats a sequence of browser steps. The immediate question is whether the assistant needs only the current page, a group of related tabs or permission to take action. Those are different requirements and should not be collapsed into a general desire for an AI browser.

Opera’s AI help page describes free use without an account, with additional history and image features associated with signing in. It also explains page context controls. Before using an assistant with a work application, understand which page content is enabled and whether the organization permits that content to be processed by the service.

Neon may be more relevant to someone who wants an agent to work across web tasks or connect an existing agent setup. A user who only needs occasional summaries may be satisfied with simpler assistance. A regulated organization may need a separate approval and deployment assessment; a consumer subscription page does not establish enterprise administrative controls or contractual terms for every use case.

03 / WorkflowPropose a research task before delegating actions

Consider a proposed evaluation comparing three publicly available software plans. This is an editorial workflow, not a hands-on test of Opera. Open the official product and pricing pages in a dedicated task or tab group. Write the questions in advance: what is included, what costs extra, which limits apply and what cannot be established from the public pages.

Begin with a summary request that explicitly asks for source links and unresolved differences. Check every numerical price and material limitation against the original page. Include a plan with annual billing and another with a monthly price to see whether the comparison preserves that distinction. A concise answer is useful only if it has not silently normalized unlike commercial terms.

Next change one source page or add a contradictory FAQ. Ask the assistant to identify the discrepancy rather than choose whichever statement is easier to summarize. This evaluates whether the browser context remains legible to the user and whether the answer distinguishes current evidence from an older description.

For a Neon agent trial, choose a harmless task such as preparing a draft comparison in a temporary document. Define the permitted pages and stop before sending a message, purchasing a plan or changing an account. Observe the proposed steps and the resulting page state. Record any clarification, correction or manual intervention needed, including cases where a page layout prevents completion.

If connecting an external agent, test that connection separately. Opera’s August 2026 announcement describes free MCP and CLI connectivity. The external agent’s own service, permissions and costs remain a separate concern. Compare the browser’s actual state with the agent’s completion report rather than treating a confident report as evidence that the requested action occurred.

04 / PricingFree assistance and paid agents have different boundaries

The current Neon subscription page lists Free and Standard. Free includes the desktop browser, basic AI chat and connections to existing agents. Standard is displayed at $19.90 on a month-to-month basis and says the price includes applicable taxes. The displayed page uses a dollar symbol; confirm the checkout currency and applicable total for your location rather than assume a currency code from the symbol alone.

Standard includes built-in agents and higher usage limits, but the page does not provide a simple numerical quota for every feature. Do not interpret a paid subscription as unlimited model or agent use. Request or inspect the limits relevant to the intended workload before making it part of a recurring process.

Opera’s August announcement confirms the free connectivity route and Windows/macOS availability. Some older wording remains inconsistent: the Neon FAQ still describes Neon as paid-only. The current subscription page and dated free-plan announcement are the stronger evidence for present plan access; verify the actual signup flow before purchase.

The Neon terms govern the paid service and explain subscription and AI-use conditions. Read the applicable cancellation and renewal terms at checkout. No subscription was purchased for this research, and no paid-model performance or completion rate was tested.

05 / DistinctionsBrowser context can remove copying without removing judgment

Opera AI’s help documentation explains that page information is sent when a user writes with page context enabled, and that the context can be switched off. That is more useful than treating the browser as an undifferentiated source of knowledge. The user needs to know which page or group is informing the response.

Neon’s FAQ describes MCP connections to a real logged-in browser session, with read-only access by default and permission needed for actions. It also distinguishes local browser interaction from cloud model reasoning. These are product descriptions to verify in the intended setup. A browser running on the user’s computer should not be assumed to mean that all AI processing stays on that computer.

Perplexity’s Blueprint offers context for source-oriented AI research. Microsoft’s Blueprint provides a broader view of assistance tied to workplace applications. Comparing these approaches helps identify whether the main benefit comes from research synthesis, existing business context or control of the browser’s active workflow.

06 / QuestionsContext exposure and completion need visible checks

Before using private pages, review the service’s data handling and the organization’s requirements. A logged-in page may contain information beyond the task’s intended subject. Use a dedicated task group and avoid unrelated sensitive tabs during an initial evaluation. This is a proposed operating practice, not a claim that a tab group is a security boundary.

For agent actions, establish what requires confirmation and how the user can stop or correct the process. The Neon FAQ describes monitoring and intervention, while the terms warn that AI services may be inaccurate and can take actions within their operating scope. Test ordinary ambiguity: two similar accounts, a changed page layout or a missing required field. The right response may be to ask for clarification.

Availability also needs product-specific checking. The current free-plan announcement supports Windows and macOS; it does not establish a mobile Neon release. Mainstream Opera AI has its own browser and platform coverage. Keep product names, operating systems and subscription requirements together in the evaluation notes so that a demonstration on one product does not become a promise about another.

07 / DecisionChoose the smallest product scope that completes the task

Opera is worth exploring when a task depends on the context already open in the browser. Start with a public research task and inspect the sources, context selection and corrections. Move to agent actions only when the expected outcome and permission boundaries are clear.

The practical choice is between free assistance, connecting an existing agent and paying for Neon’s built-in capabilities. A good evaluation records what each route actually contributes and where a person still needs to intervene. If the task is mostly reading and summarizing, action capabilities may add little. If it requires repeatable navigation and controlled changes, completion checks become at least as important as answer quality.