Skip to help content
Nuggetz
Back to Help Center

Getting Started with Nuggetz

Learn what Nuggetz does and set up your first workflow.

4 min read

Welcome to Nuggetz! This guide explains what Nuggetz does and how to set up your first workflow.

What is Nuggetz?

Nuggetz turns AI use from a solo activity into shared team play. You break work into small, ownable steps, group them into visible workflows, and let every colleague use the agents they already know without losing shared ownership, progress, or judgment.

People are the players. They own the work and make the calls. Claude, Codex, Cursor, and other agents help them execute a Nugget.

Key concepts

Nuggets

A Nugget is one small, delegable unit of work. It may repeat, but it does not have to. Each one has an owner (the person accountable for it), a trigger (what starts it), and a definition of done (how everyone knows it is finished). The definition of done is also the stop condition when an agent runs the step.

Workflows

A workflow is an ordered list of nuggets that add up to a real outcome, like publishing a weekly article or turning a lead into a booked meeting.

The three stages

Every nugget sits at one of three stages, and you move it up as trust grows:

  • Manual: a person does the step.
  • Assisted: an agent helps (drafts, prepares, suggests) and a person finishes.
  • Autonomous: an agent runs the step end to end and reports back.

Promoting a nugget to Autonomous is always your decision, and the owner can demote it any time. See Workflows & nuggets for the full picture.

Teams

Nuggetz is organized around teams. Each team has its own shared workflows, owners, agent connections, reviews, and learning. Use the team switcher to move between them.

Your first steps

  1. Describe the work. Tell Nugget Chef what you do in plain language and it proposes the nuggets, each with an owner, trigger, and definition of done. Accept the ones that fit, edit the rest.

    Prefer not to start from your own words? Pick a starter playbook instead:

    • Weekly Content Marketing Engine: plan, draft, edit, and publish content on a schedule
    • Inbound Lead → Qualified Meeting: take a new lead through to a booked meeting
    • Pull Request → Production: move a PR through review and deploy

    You can edit every nugget afterward.

  2. Set owners and definitions of done. Make sure each nugget has the right person on it and a definition of done you can actually check. This is what keeps work clear whether a person or an agent runs it.

  3. Run a Nugget with your agent. Open an Assisted or Autonomous Nugget, choose Run with…, and pick Codex or Claude Code. The first time, copy one setup command and approve Nuggetz in your browser—no API key or admin is needed. Always-on agents for scheduled work use a separate setup described in Connecting your agents.

  4. Review the first report. A nugget's first run always lands in your Review Inbox, whatever the agent reports — you see the agent's first attempt before the automation earns any trust. Approve it and the run counts as clean; correct it and the next run starts with your feedback.

Next steps