Adding Tools
Tools provide additional capabilities for an Agent to complete its assigned tasks. Some tools are pre-enabled and locked for foundational functions, while others are optional and can be enabled based on the Agent’s requirements.
Best Practice: Only enable tools that are necessary for the Agent's task. Enabling unnecessary tools may add complexity or slow response times.
Preenabled & Locked Tools
These are tools that are necessary for all Agents to function, and will always be enabled per Agent no matter what.
| Tool Name | Function |
|---|---|
| String Matcher | Finds the correct Advertiser Name, Campaign Name or other identifier based on the user’s query. If no exact match is found, the AI will ask for clarification. |
| Query Executor | Writes SQL queries to retrieve data from the assigned dataset(s). It can work with large datasets, but is limited to returning 100,000 rows or less at a time. |
| CSV Reader | Method of passing results between the Agent, its tools and the user. |
Available Optional Tools
These are tools that add to the Agent’s capabilities when enabled in order to perform its task / complete its goal.
| Tool Name | Description | Use Cases |
|---|---|---|
| Code Interpreter | Executes Python code, performs calculations, and generates charts/graphs. Required for advanced data processing and visualization of the results that were queried. | Data analysis, statistical calculations, automated reporting, chart/graph creation. |
| Image Generator | Uses Recraft to generate images. *AI isn’t perfect yet, especially in image generation; may not get you ready-to-go creative | New Logo creation, ad creatives for Facebook/Google Display, website illustrations. |
| Web Researcher | Uses Perplexity AI to research topics and summarize information from top sources. | Competitive market research, industry benchmarking. |
| Google Web Searcher | Searches Google and returns top-ranked results, using Serper.dev. | Finding up-to-date information, retrieving high-ranking web pages. |
| Web Scraper | Extracts specific content from user-provided URLs, using Firecrawl.dev. | Pulling product descriptions, pricing comparisons, available inventory. |
| Browser Use (Experimental) | Uses steel.dev to launch, view and interact with a browser in real-time *Requires ChatGPT-4o (supports vision capabilities). Slow response times expected. | Extracting images and descriptions of those images, interacting with and submitting web forms. |
Choosing the Right Tools
- If the Agent requires calculations or data visualization, enable Code Interpreter.
- If the Agent needs real-time research, enable Web Researcher or Google Web Searcher.
- If the Agent extracts data from structured websites, enable Web Scraper.
- If the Agent needs to navigate websites and interact with them, use Browser Use (Experimental).
Only activate the minimum tools needed for efficiency and faster response times.
Updated 14 days ago