Skip to content

Usage

Opera Neon is a GUI browser. All interactions happen through the browser's interface, not a terminal or SDK. The core concept is the Task - you describe what you want done and Neon's agents carry it out.

Minimal Usage Example: Research and Summarize

  1. Open Opera Neon from your Applications folder or Start menu.

  2. Open the Task panel using the task icon in the sidebar (or the keyboard shortcut shown in the browser UI).

  3. Describe your task in plain language, for example:

    Research the top 5 local SEO ranking factors for 2026 and create a summary document with sources.
  4. Neon dispatches agents - the browser assigns specialized agents (research agent, web browsing agent, document generation agent) to handle the steps in sequence.

  5. Review the output - when complete, Neon presents the finished document or result in a Neon Card or sidebar panel. You can copy, export, or continue refining with follow-up instructions.

Using MCP Connectors in a Task

If you have configured an MCP connector (for example, a local file system tool or an external API):

  1. Open Settings > Agents > MCP Connectors inside Opera Neon.

  2. Add the connector endpoint and any required auth credentials.

  3. In a new task, reference the connected tool by name:

    Pull the latest data from [your-connector-name] and generate a report comparing this week vs last week.
  4. The agent will call the MCP connector as part of the task workflow.

Tab Management via Agent

Organize my open tabs by topic and close any duplicates.

Neon's tab management agent reads your current open tabs and groups or closes them as instructed.

Key Concepts

ConceptDescription
TaskA natural-language instruction Neon executes end-to-end
Neon CardsStructured output panels where agents present results
MCP ConnectorExternal tool integration using the Model Context Protocol
CLI PathA configured terminal/CLI tool the agent can invoke during a task
SnapshotA saved state of a tab or page captured as task context

What Neon Is Not For

  • Headless/automated scraping at scale (use Playwright, Puppeteer, or Steel for that)
  • Unattended server-side browser automation
  • Programmatic control via API or SDK (no public SDK exists as of June 2026)