StackAI’s Agentic Tools give your workflows real power. Instead of just answering questions, your AI agents can now query external systems, search the web, create tickets, analyze stock data, or send emails—automatically and intelligently. In this video, you’ll learn how to add built-in tools, create custom tools using OpenAPI specs, and follow best practices to ensure your agents use the right tool at the right time.

Summary

• What is tool calling and when to use it in conversational or dynamic workflows

• Let LLMs decide what data they need, instead of hardcoding every input

• How to add tools to an LLM node in the Workflow Builder

• Browse and use pre-built tools: web search, BigQuery, Jira, Notion, LinkedIn, email, Salesforce, and more

• Example: connect a web search tool to answer real-time news queries with clickable citations

• Example: use Yahoo Finance Analytics to pull live stock charts and link to relevant reports

• Combine tools in one flow for multi-modal responses with search, charts, and docs

• Add static documents as guaranteed references using document nodes

• Build and integrate custom tools with OpenAPI/Swagger specs

• Clearly describe tool functions in your system prompt to guide the LLM

• Best practices for tool use:

  
  

  • Be specific in prompts

  • Split tasks across multiple LLM nodes

  • Include examples to guide tool selection and output format

  
  

• Learn more in the StackAI docs, including how to optimize prompts for tool usage

• Next up: continue with advanced prompt engineering and model tuning