Playbooks 101

Playbooks are the automation layer in Topo. A playbook is a flow of steps that runs on its own, triggered by a signal, a schedule or manually, without anyone sitting in a chat. If you have used a visual automation tool before, the builder will feel familiar: a canvas, nodes, conditions and branches.

Use a playbook whenever the work should happen on a recurring basis or in reaction to an event. If you want the work done now, in conversation, ask the AI Agent instead.

You will find Playbooks and Runs under Automate in the left sidebar.

Playbooks are included for Enterprise plans only.

The Playbooks page

The list has two tabs: Your playbooks and Templates.

Your playbooks shows one row per playbook, with:

Parameters at the top right holds the values and credentials your playbooks need. Columns lets you change what the table shows.

Starting from a template

Topo ships 19 templates, grouped into Sourcing, Inbound and Intelligence. Each card shows the step chips, the category, the name and a short description, for example:

• Company brief (Intelligence): research a company from its domain, write a prospecting brief and share it with the team

• Tech displacement outbound (Sourcing): when a company adopts a competing technology, craft a displacement angle

• LinkedIn engagement outreach (Sourcing): when a person engages with a LinkedIn post, craft a warm outreach angle referencing it

• Competitor customers outbound (Intelligence): scan competitor pages for customer names

The 19 templates are the same in every Topo workspace.

Click Preview flow to inspect a template before committing. The preview opens the full flow on a canvas, so you can read every step. The LinkedIn engagement outreach template, for instance, runs: Signal (client post reaction) → Enrich company → Run AI → Fit condition (is the angle relevant?) → Send notification to your team over Slack or email. For Slack to work, the Slack integration needs to be connected in Settings > Integrations first.

If it fits, click + Install in your org. Installed templates show ✓ Installed with an Open button, and land in Your playbooks where you can edit them freely to match how you work.

Creating a playbook from scratch

Click + New playbook and choose how to start:

• Blank, and build it yourself

• With AI, and describe what you want

• From a template

You can also just ask the AI Agent to build a playbook for you in chat.

Playbooks cannot be duplicated. If you need a variant, install the template again or build the new one from scratch.

The builder

The playbook editor has three areas.

Add a step (left panel) is the node library, searchable and grouped:

The canvas (centre) is where the flow lives. Zoom controls, fit-to-screen and a {} view for the underlying definition sit along the bottom, next to a live count of errors and warnings. A playbook with errors will not publish, so work that count down to zero.

Playbook setup (right panel) holds the three things that make a playbook safe to run:

• Trigger — when this playbook starts. On a signal trigger, Starts on is either Any signal of a type or Specific signals. Specific signals is a multi-select: pick several signals you already set up, and the playbook starts when any one of them fires. The start node then reads One of N specific signals fires. A playbook that was bound to a single signal keeps working, and shows up in that same list.

• Parameters — the values and credentials the playbook needs

• Rules & limits — approvals, quotas and run limits

Along the top you have an AI assistant to edit the flow by chatting, undo and redo, version history, and Publish. There is no separate Test mode and no Test run button in the header. The header pill shows where the playbook stands:

• Draft · saved — never published

• Published · off — it has a published version, and it is not running

• Live — it is on and will run when the trigger fires

• Paused — it was on, and you paused it

Unpublished edits on a live playbook still show as unpublished changes until you publish again.

Testing a playbook

Testing happens in the editor, on the same canvas you edit.

• A bar at the bottom of the canvas shows the selected test case and Run test. Run test runs every case. The case menu shows how many passed.

• Add a case from that bar: Replay a past run, Generate with AI, or Start from scratch. When the trigger is a signal, the menu also has From a signal. It copies a recent imported occurrence that matches the trigger, and the case is described as From a real signal. On a signal's Imports log, an imported event has Test in a playbook.

• Open a step and use its Output tab. During tests is what that step does in a test (simulate, skip, send to you only, run for real, and so on). The same tab shows the last result for the selected case. Test this step runs the full test on the selected case.

• Before a run, the canvas only badges a step whose test setting differs from the default. After a run you see the result on each step and the path that ran.

• Any change to the playbook, a test setting, a pinned output or a case clears the results. Test again to refresh them.

• A loop runs its first 3 items in a test. A wait does not wait, so later steps run straight away. Live runs are unchanged: they run every item and they wait.

Test runs are also listed on the Runs page, tagged Test run.

Staying in control of what a playbook does

A playbook only contacts leads if you build it that way. Nothing goes out unless a step in the flow sends it, so a playbook that enriches a company and notifies your team on Slack will never message anybody.

When a flow does reach leads, and you want a human check before it does, add an Approval step. The run pauses there until someone approves it.

Any member of your workspace can publish a playbook. Publishing saves a version. It does not start the playbook. If it is not already running, the toast says Playbook published and It won't run until you turn it on, with a Turn on action. On the list, that playbook shows Off until you turn it on, and Turn on works without asking you to publish again.

Runs themselves do not consume credits: only the steps that enrich leads do, exactly as they would if you ran the enrichment by hand.

Recommended way to ship one

1. Start from a template if one is close, or ask the AI assistant to draft the flow.

2. Set the Trigger and fill in Parameters. If the same flow should follow several signals, pick all of them under Specific signals instead of building one playbook per signal.

3. Set Rules & limits before the first live run. Add an Approval step if the playbook sends anything to a lead.

4. Use the test bar to Run test, and read the result on the canvas. Test runs also appear on the Runs page.

5. Clear all errors, then Publish. If the playbook is not running, choose Turn on.

Checking what ran: the Runs page

Runs in the sidebar lists every run, test runs included:

Search by playbook name, or use Filters and Columns to narrow the view. Click the arrow on any row to open the run and inspect it step by step, as you would in any automation platform.

Still have a question? See FAQ: AI Agent, Memory, Skills & Playbooks.