Runbooks your teamcan follow.

A playbook walks a person through an operational task, like handling a refund or escalating a customer-reported bug to Engineering. Each step can call a tool in your connected apps, check the result, and pass values on to the next step.

Handle a refund request5 steps · support
  1. 1
    Find the ordercollect

    Order ID

    ORD-10482

  2. 2
    Check it’s eligiblecheck

    refunds-admin · check_refund

    eligible is true · days_since_payment is at most 30

  3. 3
    Pick the refund typedecide

    Full, partial, or store credit

    Full refund

  4. 4
    Issue the refundaction

    refunds-admin · issue_refund · runs as you

    Output · Refund RF-2291

  5. 5
    Confirm the money movedverify

    refunds-admin · refund_status

    status was pending, expected processing or settled · recorded, you can go on

Follow-up · check refund_status again until it is settled
An example run, laid out from the playbook’s steps and results.

Steps that use
the tools you have.

Connect the apps your team already works in through MCP connectors. A step can call one of their tools, like creating the Jira ticket at the end of an escalation.

  • Jira
  • Confluence
  • Linear
  • GitHub
  • Notion
  • Sentry
  • ElevenLabs
  • and more
  • Each member signs in to their own account with OAuth, so tool calls run as the person running the playbook.

  • Connector secrets are stored encrypted. The same connectors are available in chat.

escalate-bug.playbook.yaml
tools:
  create-ticket:
    connection: jira
    tool: createJiraIssue
    title: Create the Jira ticket

steps:
  - title: Escalate to Engineering
    type: action
    action:
      tool: create-ticket
    outputs:
      ticket:
        field: key
        label: Jira ticket

Checks that advise.
They never block.

A failed check doesn’t stop the person running the playbook. The problem is recorded on the run and shown, so someone can follow it up.

expect

Compare the result

A step’s check compares what a tool returned against the values you expect, field by field.

outputs

Carry values forward

Keep a value from a result, like a ticket ID, and use it in later steps and in the run summary.

follow-up

Check again later

Re-check after the run, for example until a refund settles. Follow-ups run in the background only for connectors that use workspace credentials.

Start from what
you’ve already written.

Draft with AI

Describe the task, or point at an existing doc such as a Confluence page. AI drafts the playbook for you to review and edit.

A video of every playbook

Each step becomes a scene with narration, in your workspace’s video style. New people can watch the whole task before they run it.

About walkthrough videos

Stop and pick up later

Stop a run while you wait on someone, and resume it later from where you left off.

Turn a doc into
a playbook.

Pick a task your team repeats, connect the apps it touches, and run it step by step.