Skip to main content
Playbook Playbooks are Voiceflow’s core building block for agentic AI. Each Playbook contains instructions that guide how your agent responds to users, plus access to tools it can use to take actions like searching your knowledge base, querying external services, or sending messages with buttons and cards. Unlike Workflows, which follow a fixed sequence of steps, Playbooks let your agent reason through conversations and decide how to respond based on context. Use Playbooks when you want your agent to handle open-ended requests without manually defining every possible path. Playbooks can be called directly by your Agent, or from within a Workflow using the Playbook step. Multiple Playbooks in a workflow can communicate with each other using a Crew step.

​
Creating a Playbook

Playbooks can be created from within the Agent tab by pressing the + button next to the Playbooks header. Each Playbook has a name and a set of instructions that define its behavior. You can also create Playbooks via the Playbook step, or by pressing Playbooks → New Playbook on the sidebar.

​
Writing instructions

Instructions are the core of every Playbook. They tell your agent what it should do, how it should behave, and when it should use its available tools. Write instructions in natural language, using markdown formatting if needed. Good instructions are specific about your agent’s role and responsibilities. Instead of writing “help users with support issues,” specify the types of issues the agent should handle, how it should gather information, when it should escalate, and what tone it should use. Here’s an example of some well-written instructions: 
TODO
TODO
TODO
TODO
You can edit instructions manually or click the AI button to open an editor that helps you refine your prompt.

​
Configuring model settings

Click Model in the top right corner of the Playbook editor to configure the AI model your Playbook uses. You can adjust three settings:
  • AI model: Choose which model powers your Playbook (eg: Claude 4 - Sonnet).
  • Temperature: Control how creative or deterministic the responses are. Lower values produce more consistent responses, while higher values allow more variation.
  • Max tokens: Set the maximum length of the model’s response.
By default, your Playbook will use the same model settings as your Agent.

​
Adding tools

The Tools panel on the right side of the Playbook editor shows all the tools your agent can use. Click the + button to add integration tools like Zendesk, Salesforce, or Gmail, or custom tools like the API tool or Function tool. You can also toggle System tools from the same sidebar. System tools let you enable built-in capabilities such as searching the knowledge base, sending visual artifacts like cards, or searching the web. When a system tool is enabled, your agent can decide when to use it based on the conversation context and your instructions.

​
Configuring exit conditions

Exit conditions are only available when a Playbook is used via a Playbook step in a workflow. When a Playbook is triggered by your Agent, exit conditions will be ignored.
Exit conditions let your Playbook hand off to other parts of a workflow when specific situations arise. For example, you might create an exit condition that triggers when a user wants to speak with a human, then route that conversation to a Call forward step. Click Exit conditions at the bottom of the Playbook editor to add conditions. Each exit condition needs a name and a description that helps your agent understand when to trigger it.