LLM agents can be equipped with a wide variety of tools, depending on their use case and domain. These tools allow agents to go beyond simple language generation and take real actions, retrieve data, or interface with external systems.

The most common categories of tools include search, math, code execution, file I/O, API calling, and domain-specific utilities.

🔎 1. Search Tools

• Web Search APIs: SerpAPI, Google Search, Bing Search
• RAG Retrieval Tools: VectorDB queries using FAISS, Chroma, Pinecone

➕ 2. Math and Computation Tools

• WolframAlpha: External knowledge and symbolic math
• llm-math: LangChain’s built-in math tool using Python evaluation

🧠 3. Code Execution Tools

• Python REPL: For local computation, logic evaluation, and loops
• Code interpreter / Advanced Data Analysis (ADA): OpenAI’s sandbox for running Python securely

📁 4. File & Document Tools

• ReadPDF: Parse and extract content from uploaded PDFs
• CSV Analyzer: Process spreadsheet-like data files
• File Search: Locate information from a corpus or folder

🌐 5. Web & API Access Tools

• HTTP Request: Generic GET/POST/PUT tool for REST APIs
• GraphQL Tool: Schema-based API calling for structured data services

🏢 6. Domain-Specific Tools

• CRM Connectors: Salesforce, HubSpot, Zoho integrations
• Financial Data Tools: Yahoo Finance, Alpha Vantage
• IoT Device Control: Home Assistant APIs, robotic controls

🧰 Sample Tool Call

{
  "tool": "getStockPrice",
  "input": {
    "symbol": "AAPL"
  }
}

📦 Toolkits in Frameworks

• LangChain: Includes built-in toolkits for web search, math, shell, python, and vector DBs
• AutoGPT & AgentLabs: Use modular tool plugins and environment connectors

🚀 Summary

Tools bring the power of action and perception to LLM agents. By integrating common tools like search engines, math solvers, file readers, and external APIs, you enable your agents to work across domains, perform real tasks, and adapt to evolving information and user needs.