Agent overview

The Endstack agent is the AI that works inside your Endstack workspace. It can read channels and use tools on your desktop.

You can also contact the agent through Slack, by text message, or by emailing it at its Endstack email address after you set up those integrations. For mobile use cases, we recommend using the Endstack mobile app, which offers a much more complete experience than texting the agent.

What the agent can do

The agent can:

• Reply in channels.
• Use built-in tool groups for the filesystem, Chrome, email, settings and wallpaper, apps and windows, X (Twitter), and more.
• Load skills, which provide reusable instructions and workflows.
• Call tools from MCP servers that you configure.
• Use connected app integrations.
• Run on a schedule through task channels.
• Respond to remote triggers through email, messaging, and Slack.

Every run operates within the access of its trigger user. The agent can use that person's resources and shared workspace resources, but it cannot silently access another member's private files or credentials. See Autonomy and approvals for the full resource boundary.

Where runs appear

Agent activity appears in the channel where it was triggered. While a run is active, the channel can show streaming status and tool-approval cards with Approve and Deny actions. The completed response appears in the same channel.

The dock can also indicate that an agent run is active. Scheduled runs appear in their task channels, while conversations started through a first-party integration typically appear as external channels.

Configure the agent

Per-channel behavior—including context Markdown, proactive-reply controls, and task tool whitelists—is configured separately. See Channels and Channel settings.

Related pages

• Autonomy and approvals
• Tools
• Models
• First-party integrations