Skip to main content

HERE MCP get-browser-context tool

get-browser-context reads the text from web pages the user has open in Enterprise Browser (EB), so your agent can provide responses relevant to the user's current work.

ToolPurpose
get-browser-contextRead text and markdown from the user's open EB views

New to the MCP server? Start with the overview.

What it does​

get-browser-context returns text and markdown pulled from the user's active EB views. Your agent can use it to ground answers, summaries, or actions in what's on screen.

A typical flow:

  1. Call get-browser-context to see what the user is working on.
  2. Have your agent reason over that content.
  3. If needed, act on it with raise-intent or send-notification.

To find out which window, supertab, and view the user is focused on (rather than the text inside them), use get-focused-content.

Security model​

Admins control which content the tool can read:

  • Views are only included if AI Participation is enabled for the app. Apps with it disabled are skipped. See Additional Content Controls in Manage apps.
  • Entitlements still apply. Your agent can't read content from apps the user can't access.
  • The tool always reads for the subject (sub) of the bearer token.

HERE truncates the text on the server so responses fit typical agent token budgets.

Prerequisites​

  1. You can connect to the MCP server. See Connecting for the endpoint and authentication. Then call tools/list to check that get-browser-context is available and see its current input schema.
  2. AI Participation is enabled for the apps you want agents to read.
  3. Enterprise Browser is running with at least one eligible view open. If no eligible views are open, you receive little or no content. It doesn't generate content to fill the gap.

Example​

const context = await mcp.call("get-browser-context", {});
// Use the returned text and markdown to ground the agent's next response

Limits​

  • Eligible views only. You never receive content from apps with AI Participation disabled.
  • Long pages are truncated. The content you receive might not include the full page.
  • Live session only. The tool reads the user's current EB session, not their browsing history.
  • Current user only. The tool only reads for the authenticated user.

Reference​

Table of Contents