Skip to main content
Playbooks main view Playbooks are AI-powered workflows that automate complex tasks using natural language instructions. They execute using Claude AI and can leverage any MCP servers installed in your Toolbox.

​
What are Playbooks?

Playbooks allow you to:
  • Automate workflows with natural language instructions
  • Use MCP tools installed in your Toolbox to access external APIs and services
  • Track execution history with detailed logs and results
  • Schedule runs or trigger them manually
  • Configure AI parameters like model, temperature, and timeouts

​
Creating a Playbook

1

Navigate to Playbooks

Click “Playbooks” in the sidebar navigation
2

Click New Playbook

Click the “New Playbook” button in the top right
3

Enter Playbook Details

Provide a name and description for your playbook
4

Write Instructions

Write natural language instructions describing what the playbook should do
5

Save

Click “Create” to save your playbook
Playbooks have access to all MCP servers installed in your Toolbox. Make sure to install the necessary servers first!

​
Playbook Instructions

Playbook instructions Instructions are written in natural language and tell the AI what to do. Be specific about:
  • What data to gather: Which APIs or tools to use
  • How to process it: Any transformations or analysis needed
  • What to output: The format and content of the result

​
Example Instructions

You are a stock advisor given the reports provided by Finnhub.
You ingest articles from Finnhub and recommend the top 3 stocks
expected to provide the greatest returns over the next week.
Natural Language Instructions: Write instructions as if you were explaining the task to a human. The AI will interpret and execute them using available tools.

​
Playbook Configuration

Playbook configuration Configure how your playbook executes:
​
Status
toggle
Enable or disable the playbook. Disabled playbooks cannot be executed.
​
Model
select
The Claude model to use for execution (e.g., claude-sonnet-4-20250514)
​
Temperature
number
Controls randomness (0.0-1.0). Lower values are more focused and deterministic.
  • 0.0-0.3: Precise, factual tasks
  • 0.4-0.7: Balanced (default: 0.7)
  • 0.8-1.0: Creative, exploratory tasks
​
Max Tokens
number
Maximum tokens for the response (default: 4096)
​
Timeout
number
Maximum execution time in seconds (default: 300)

​
Running a Playbook

1

Open Playbook

Click on a playbook from the list
2

Click Run

Click the green “Run” button in the top right
3

Monitor Execution

Watch the execution progress in real-time
4

View Results

See the execution result and any tool calls made
Playbooks execute asynchronously. You can navigate away and check the results later in the Recent Executions tab.

​
Recent Executions

Playbook recent executions The Recent Executions tab shows a history of all playbook runs:
  • Execution status: Completed, failed, or in progress
  • Timestamp: When the playbook was executed
  • Duration: How long the execution took
  • Executor: Who triggered the execution
  • Detailed steps: Expand to see reasoning, tool calls, and results

​
Execution Details

Each execution shows:
The user who triggered the execution and their email address
When the execution started and how long it took to complete
The natural language instructions that were executed
The final output from the AI, including any recommendations or findings
A detailed breakdown of each step:
  • Reasoning: The AI’s thought process
  • Tool Calls: Which MCP tools were invoked
  • Results: The output from each step

​
Playbook Status

Playbooks can have the following statuses:

🟢 Active

Playbook is enabled and can be executed

⚫ Inactive

Playbook is disabled and cannot be executed

​
Playbook Management

​
Edit a Playbook

  1. Click on the playbook to open it
  2. Navigate to the Instructions or Configuration tab
  3. Make your changes
  4. Changes are saved automatically

​
Delete a Playbook

  1. Click on the playbook to open it
  2. Click the menu icon (three dots) in the top right
  3. Select “Delete Playbook”
  4. Confirm deletion
Deleting a playbook will also delete all execution history. This action cannot be undone.

​
Best Practices

Be Specific

Write clear, detailed instructions about what you want the playbook to do

Install Tools First

Make sure required MCP servers are installed in your Toolbox before creating playbooks

Test Thoroughly

Run your playbook multiple times to ensure consistent results

Monitor Executions

Check execution logs to understand what tools were called and why

​
Common Use Cases

Gather data from APIs, analyze trends, and generate insights
Collect information from multiple sources and compile reports
Check systems or data sources and notify when conditions are met
Search multiple databases and summarize findings

Toolbox

Install MCP servers to give your playbooks new capabilities

Activity Logs

Track all playbook executions across your workspace

Workspace Management

Manage workspaces to organize your playbooks

API Tokens

Generate tokens for programmatic playbook execution
⌘I