Platform Actions Building Blocks for AI Abilities

Actions are the executable components within skillset abilities that perform specific tasks. Each action is specified using a markdown-style code block, where you specify the action name after the first 3 backticks.

  • search

    Searches the web or a dataset with a specific phrase
    @search/web
    @search/news
    @search/images
    examples
  • fetch

    Fetches a URL with built-in timeout and retry logic
    @serper/web/search
    @brave/web/search
    @tavily/search
    examples
  • email

    Sends an email to a specified recipient
    @email/send
    examples
  • echo

    Echoes back a message or value
  • abort

    Aborts the current operation or workflow
  • view

    Uses vision model to describe an image from external URL
    @view/describe
    examples
  • listen

    Listens for user input or specific events
  • blueprint

    References a blueprint and its associated resources
  • bot

    References a conversational bot configuration
  • dataset

    References a dataset for searching or data retrieval
  • skillset

    References a skillset with specific abilities
  • memory

    Accesses stored information or conversation context
  • space

    References a workspace or environment configuration
  • file

    References or processes a file resource
  • attachment

    Handles an attached file or resource in conversation
  • text

    Generates text using input text as a prompt
    @text/generate
    @text/summarize
    @text/translate
    examples
  • image

    Generates an image using input text as a prompt
    @image/generate
    @image/generate[dalle3]
    @image/generate[gpt-image-1]
    examples
  • form

    Processes structured data entry from a form
  • shell

    Executes a shell command or script
    @shell/exec
    @shell/read
    @shell/write
    examples
  • browser

    Performs a web browser automation action
  • conversation

    References or manages a conversation thread
  • task

    Creates or manages a task or to-do item
  • pack

    Defines a collection of multiple abilities grouped together
    @pack/vanta
    @pack/vanta[vendor]
    examples
  • agent

    Executes a task using an AI agent with specific instructions
    @perplexity/search[sonar]
    @agent/task/evaluate
    @agent/task/plan
    examples
  • mcp

    Dynamically loads Model Context Protocol tools from external servers
    @mcp/load[notion]
    @mcp/load[linear]
    @mcp/load[box]
    examples
  • todo

    Manages a temporary todo list stored in Redis with automatic expiration
    @todo/read
    @todo/write
    @todo/manage
    examples

Frequently Asked Questions

What is an action?

An action is an executable component within skillset abilities that performs a specific task. Actions include things like fetching data from URLs, generating images, managing files, sending emails, and more. They are specified using markdown-style code blocks with the action name.

How do I use actions?

Actions are used within skillset abilities by creating a code block with the action name. For example, to fetch data from a URL, you would create a code block with "fetch" and specify the URL and other parameters in YAML format.

What actions are available?

ChatBotKit provides a wide range of actions including search, fetch, email, text, image, file, shell, mcp, and many more. Each action is designed to perform a specific task that can be integrated into your AI workflows.

Can I combine multiple actions?

Yes, you can combine multiple actions within a single ability to create complex workflows. Actions can be chained together and mixed with free-form text instructions to guide how they should be applied.

What is the MCP action?

The MCP (Model Context Protocol) action dynamically loads tools from external MCP servers. This allows you to extend your AI capabilities by connecting to services like Notion, Linear, Box, and other MCP-compatible tools.