Endstack

Tools

Explore the built-in capabilities the Endstack agent can use across your desktop, browser, files, and connected services.

Tools let the agent act inside your desktop instead of only returning text. Built-in tools cover files, Chrome, email, settings, search, apps, windows, and other endOS surfaces.

There are no per-tool toggles in Settings. Access is governed by autonomy and approvals, optional task tool whitelists, and whether a required integration is connected.

Where tools appear

During a run, you may see:

  • Tool-approval cards in the channel when a permission gate applies.
  • Streaming or compact agent status while a tool is active.
  • Results produced through skills, MCP, or app integrations.

Built-in tool groups

File System

The agent can create, read, edit, move, rename, and delete files and directories. It can also analyze images from local files, screenshots, or remote URLs.

File access remains subject to the run's user resource boundary. See Autonomy and approvals and Filesystem.

Chrome Browser

The agent can manage Chrome profiles, windows, and tabs; navigate to pages; inspect page content; and interact with elements by clicking, typing, scrolling, selecting, or hovering.

Chrome tools can browse any website, including interactive and authenticated pages available in the browser. See Search for when the agent uses Chrome instead of a search API.

Email

The agent can list inbox messages and threads, read messages, send new mail, reply, and reply-all from the instance inbox. These tools use the built-in Endstack inbox described in Email.

Settings and Wallpaper

The agent can read and update desktop, theme, terminal, notification, agent, and layout settings. It can also manage theme presets and set wallpaper from a file or URL.

See Appearance for the corresponding user-facing controls.

The agent can use live web search through Exa's web_search, structured X/Twitter research tools, and Chrome for interactive browsing. See Search for the capabilities and typical use of each option.

X (Twitter)

X tools can search posts, fetch a post's context and replies, count posts over time, retrieve trends, and load news stories. These tools are covered in detail in Search.

Apps and Windows

The agent can discover installed apps, list open windows, and control windows. Available actions include opening, focusing, capturing a screenshot or snapshot, and closing a window.

MCP tools

Connected MCP servers expose additional capabilities, typically with generated names beginning with mcp_*. Server configuration and workspace sharing controls are documented under MCP.

Extension surfaces

Several related capabilities extend what a run can do without appearing as separate per-tool Settings pages:

  • Skills provide packaged workflows and instructions. See Skills.
  • App integrations provide toolkits for connected third-party accounts. See App integrations.
  • Tasks provide recurring or scheduled runs. See Tasks.
  • Messaging and Slack enable remote conversations when their first-party integrations are enabled. See Messaging and Slack.
Documentation