Skip to main content

Designing your first Agent

Agents in Coworker are AI-powered assistants that can handle specific, recurring tasks on your behalf — automatically, on a schedule, or on demand.

Updated this week

What is Agent Builder?

Agent Builder lets you create a custom agent by writing plain-language instructions — no coding required. You give the agent a goal, connect it to the relevant tools and data sources, and decide how it gets triggered. Once set up, your agent can run autonomously and deliver results directly to you via Slack, email, or the Coworker interface.

When is an Agent useful?

Agents work best for tasks that are repetitive, time-consuming, or require pulling information from multiple places. A few common examples:

  • Daily or weekly digests — Automatically summarize Slack activity, emails, or meeting notes and deliver a briefing each morning

  • Pipeline monitoring — Scan your CRM for stale deals and send a Slack alert when action is needed

  • Status reports — Pull data from Jira, GitHub, or your project tools and generate a structured update on a schedule

  • Customer health tracking — Monitor activity across accounts and surface at-risk customers before issues escalate

  • Document or inbox triage — Process incoming emails, invoices, or documents and route them appropriately

ℹ If you find yourself doing the same research or reporting task more than once a week, it's a good candidate for an agent.

How to build your first Agent

  1. Go to Agents in your Coworker workspace and click Create New Agent

  2. Name your agent with a clear, descriptive label (e.g., "Weekly Pipeline Digest" or "Morning Slack Summary")

  3. Provide your instructions — describe the task in plain language. Be specific about what data to use, what to look for, and how to format the output

  4. Select tools and data sources — grant access only to what the agent needs (e.g., HubSpot for deal data, Slack for messages)

  5. Set a trigger — choose manual, scheduled (e.g., every Monday at 9am), or API-based triggering

  6. Save and test — run the agent and review its output. Adjust your instructions as needed

Tip: Start simple. Build a single-task agent first, confirm it works reliably, then expand its scope from there.

Did this answer your question?