Appearance

Creating & Editing Runbooks ​

Runbooks are authored in an assistant-first editor: the steps you're writing sit on the left, and an AI assistant that can draft and refine them sits alongside on the right. Outside of editing, every runbook opens in a locked read view so procedures can't be changed accidentally.

You can write runbooks by hand, but the fastest — and usually most accurate — way is to let the agent build them with you. There are two ways to do it:

From a Real Ticket in Explore ​

The best runbooks capture what actually worked, and the General Assistant (Explore) can turn a resolved ticket into a runbook for you:

  1. Open the General Assistant and pull up a ticket your team resolved — "Show me ticket #12345 and walk me through what was done."
  2. Review the steps and tools that were used to resolve it.
  3. Ask the agent to build a runbook from it — "Create a runbook for handling this type of issue, based on how this ticket was resolved."

Because the runbook is grounded in a real resolution — the actual steps taken and the actual integration tools used — it tends to match future tickets better than a procedure written from memory. Attach the source ticket under Applies to tickets to strengthen matching further.

The agent can manage the full runbook lifecycle from the conversation: create new runbooks, update existing ones, publish them, and archive runbooks you've retired. You always get to review what it produced — open the runbook in the editor to refine before it goes live.

With the Assistant in the Editor ​

The Assistant panel is open alongside the steps whenever you edit a runbook. Describe the procedure — or paste rough notes — and the assistant drafts and refines the steps in real time as you chat. Try prompts like "Add a troubleshooting section" or "Improve the formatting." Templates use the same assistant, seeding it with a starter prompt for the procedure you picked.

Creating a New Runbook ​

Navigate to Runbooks in the sidebar and click New Runbook. Choose the runbook's Scope — Global for all companies, or Company-Specific for a single client (then pick the company) — and click Create.

You land directly in the editor with the Assistant panel open. Give the runbook a title, then write the steps yourself or ask the assistant to draft them.

Scope is fixed at creation

A runbook's scope is set when it's created (or forked) and can't be changed in the editor. To get the same procedure at a different scope, fork it.

The Editor ​

The editor has two panes:

  • Left pane -- The runbook itself: title, the Auto-run switch, the Applies to tickets card, a Tools used row, and the Steps editor.
  • Right pane -- The Assistant panel. Collapse it with the chevron when you want the full width; an Assistant button brings it back.

Steps ​

Steps are written in a block-based editor — type / for commands and # to reference a tool. Supported content blocks:

  • Headings for organizing sections.
  • Paragraphs for descriptive text and instructions.
  • Bullet and numbered lists for step-by-step procedures.
  • Images uploaded directly into the editor.
  • Mermaid diagrams for flowcharts and decision trees (type /mermaid).
  • Code blocks for scripts and commands.
  • Wait blocks that pause an execution at a specific point — the run moves to Paused runs until a technician follows up. See Executions & Monitoring.

If the runbook needs information before it can run — a user's email, a device name — document it in the steps. The agent asks for anything it's missing when an execution starts.

Tool References ​

Type # in a step and pick an integration tool by name to reference it. Every tool referenced in the steps is detected automatically and shown in the Tools used row, so you can see at a glance which integrations a runbook depends on.

Saving ​

Click Save changes to persist everything — title, steps, tools, and auto-run settings — in one action, or Cancel to discard your edits. After you save a new draft for the first time, Junto asks "Ready to publish?" — publish right away, or choose Not yet and keep refining.

The Read View ​

Outside the editor, every runbook opens locked and read-only — the status badge at the top shows Locked · read only. This is by design: published procedures can't drift from casual browsing, and executions always run the published version's content, so in-progress draft edits never affect live runs.

From the read view you can:

  • Edit -- Unlock the runbook in the editor.
  • Run now -- Execute the runbook immediately (published runbooks only).
  • Schedule -- Put it on a recurring cadence (published runbooks only). See Runbook Scheduling.
  • Runs -- Review this runbook's execution history.
  • Fork -- Copy it to a different scope.
  • Publish -- Publish a draft so it can run.
  • The ⋯ menu holds Export (download as Markdown) and Delete (archives the runbook).

Approvals First: Building Trust ​

A new runbook doesn't have to start fully autonomous. Until you enable auto-run, a matched runbook runs supervised: each action the agent wants to take is sent as an approval request to the Slack or Teams channel you've routed to the Runbook Approvals slot (see Notification Channels), and you approve or deny every step without leaving Slack or Teams.

This is the right mode for new runbooks and higher-risk procedures — you see exactly what the runbook does, action by action, while it earns your trust.

Auto-Run ​

When a runbook has proven itself — and you're tired of clicking Approve — turn on auto-run. The runbook then executes end to end, with no technician kickoff and no per-action prompts, whenever triage matches it to an incoming ticket with high confidence.

Flip the Auto-run switch in the editor and configure:

  • Runs unattended as -- The technician who gets credit for the run. Ticket updates and time entries in your PSA are attributed to this user. Required when auto-run is enabled.
  • Min confidence -- The runbook only auto-runs when the triage match confidence meets or exceeds this percentage (50–99%, default 75%). Below the threshold, the runbook is recommended but not executed. The slider runs from More runs to Fewer, safer — start toward safer and lower it as you build trust in the runbook's matching accuracy.

Auto-run settings are saved with Save changes. Auto-run is available to users with the Manager role or above, and only when auto-run is enabled for your organization.

Applies to Tickets ​

To improve how reliably triage matches a runbook to the right tickets, attach real tickets that this runbook should handle in the Applies to tickets card:

  1. In the editor, search tickets by ID or description and attach the ones that represent this runbook's use case.
  2. Remove tickets that no longer represent the runbook.

When the runbook completes a successful execution, the source ticket is attached automatically — so matching gets better the more the runbook runs.

Forking ​

Use Fork to copy a runbook to a different scope — for example, a company-specific variant of a global procedure — without affecting the original. Choose the new scope (and company, if company-specific) and click Create fork.

Managing the Runbook List ​

The Runbooks page includes tools for keeping a large library organized:

  • Search and filters -- Search by title, description, or tags, and filter by Category, Scope, Status, Company, or Integration (which runbooks use tools from a given integration).
  • Bulk actions -- Select multiple runbooks to Publish, Unpublish, or Delete them at once.
  • Schedule indicators -- Runbooks with active schedules show a Scheduled badge (with a count); paused schedules show Paused.
  • Archived runbooks -- Hidden by default; use Show Archived to display them.

Next Steps ​