AI Agents configuration

the ai agents configuration tab, located on the left sidebar, is where you can create and configure ai agents this article serves as a reference for all fields in this tab system prompt the system prompt is your agent's core instructions it outlines its purpose, behavior, and any guidelines or constraints it should follow the agent follows these instructions across all workflows the improve button on the lower right hand side is an optional prompt improvement feature use it to improve your existing prompt with ai and enable more consistent outcomes after you click improve , you can leave the field empty for an auto improvement or suggest specific improvements click save to apply the changes context in context , you can upload external information to improve the agent's knowledge base examples include internal knowledge bases or reference tables {{product name}} stores the files in context in a rag database, enabling long term ai memory current limitations for uploading in context txt, pdf, docx, csv, and json files only 20 mb maximum per file 50 files maximum per team 100 files maximum per organization 20 files maximum per ai agent you can also view your uploaded context files in the make ai agents app in the {{scenario singular}} builder mcp in mcp , you can connect your agent to mcp servers to access additional tools click the add button to create a connection to an available mcp server, then toggle on the tools that you want the agent to access click save tools tools are {{scenario plural lowercase}} your agent uses to make decisions and complete tasks these are tools the agent can always access across all workflows changing system tools impacts all {{scenario plural lowercase}} using an agent when an agent processes requests, it uses tool names and descriptions to choose the right tool for the task similarly, it uses scenario inputs and outputs docid 7odicud0ke9 tr4xwuiqn names and descriptions to choose the right data testing & training testing & training is a chat interface that enables direct communication with your agent, eliminating the need to connect to external tools like telegram or slack use it to test, debug, and refine your agent's responses when you interact with your agent in testing & training , the agent can show you the tools it uses and their inputs and outputs you can click the new chat button to start a new chat for reference information on the make ai agents app, see make ai agents app docid\ mdyjgeeiirrxmd607v1xg