AI agents are intelligent systems designed to process information, make decisions, and execute tasks on behalf of users. They can automate workflows, enhance productivity, and integrate seamlessly with various tools and applications. One significant advantage of Scade AI agents is that you can create any required app from 1500+ available multimodal nodes and use it as a tool within your agent, solving your specific problem efficiently.
To get started, you can either try one of the existing Scade agents or create a new one.
Creating an AI Agent
Navigate to the Agents Tab: Go to the "Agents" tab on the Scade platform.
Press the Create Button: Located at the top-right corner.
Provide a Name and Description: Define what the agent is for.
Setting up a system prompt
💡System prompt is like a main instruction for an AI agent, where we explain, how our AI agent should approach whenever we give him a task.
PRO TIP: use our Agent Prompt Generator to help you draft your system prompt
===ROLE===
You are an AI agent responsible for creating and posting X threads daily based on topics from the Acme Labs LLC annual report.
=== INSTRUCTIONS ===
1. Use the knowledge base to access and review the Acme Labs LLC annual report.
2. Identify key topics and insights from the report that can be used to create engaging threads.
3. Compose X threads daily, ensuring each thread is informative and relevant to the identified topics.
4. Use the X posting tool to schedule and post the threads at optimal times for maximum engagement.
5. Maintain a professional and informative tone, ensuring clarity and accuracy in each thread.
=== IMPORTANT RULES ===
1. Reply in the language used by the user in their message.
2. Do not overload the system with multiple requests simultaneously; handle one task at a time.
3. Ensure all threads are based on verified information from the annual report in the knowledge base.
If following a step-by-step approach, ensure that the agent adheres to each step strictly. You may also specify whether you require confirmation at each stage.
If using a one-step approach, state the expected result and tasks clearly before execution.
When referring to the knowledge base, use phrases like:
"Go to..."
"Check..."
"Find out..."
When referring to Subagents use words like communicate, explain, and talk.
When invoking tools, refer to them using their names in quotes, e.g., "tool-name" tool.
If running a tool multiple times, define the order and specific tasks for clarity.
If a tool returns an image, request the agent to provide the image URL to avoid potential bugs.
Advanced settings
💡 For small agents we recommend stick with gpt-4o-mini, since it is the fastest model. But if you struggle with output result or require some complicated logic, choose gpt-4o or deepseek-r1.
In the advanced settings, you can:
Choose the desired LLM model, such as
gpt-4o-miniordeepseek-r1.Add a suggested prompt example to guide users.
Define how the agent will be used effectively in different scenarios.
Knowledge base
The knowledge base provides essential information that the agent uses to make informed decisions and perform tasks. Supported formats include:
Documents: md, html, pdf, doc, docx
Spreadsheets: xls, xlsx
Presentations: ppt, pptx
Text files: txt, csv, and more.
Tools
Tools are categorized into two groups:
Toolkit: Preinstalled connectors such as Google Trends, Google Drive, Gmail, Twitter, YouTube, and web search.
Your Apps: Custom-built and published workflows.
💡 For scraping Facebook, Twitter, YouTube videos, and more, we recommend creating an AI app with Apify Node, as it provides robust and scalable automation. Just keep in mind that you’ll need to use your own credentials for Apify to access the data.
When publishing a workflow:
Provide a clear description so the agent can understand its purpose.
If you make changes, update the published app to ensure the latest version is available.
Subagents
Subagents act as secondary agents that the main agent can delegate tasks to, facilitating collaboration. Key principles:
If subagent actions require confirmation, the main agent should ensure that the user approves each step before proceeding. Interim results should be recorded before arriving at a final outcome.
Automation
Automate your AI agent by scheduling predefined scenarios or integrating with external services for continuous operation.
Available automation options:
Cron(time-based scheduling) andTelegram.
API integration
Your AI agent can be integrated into applications via API. The API follows a standard chat completions format with support for roles:
usersystemassistant
Pre-configured API request samples are available in:
CurlJavaScriptPython
Input and output schema
Defining an input-output schema ensures structured interactions with the agent.
Input fields: Standardized formats for incoming requests.
Output fields: Well-defined responses that external systems can interpret correctly.
This structured approach ensures seamless communication and integration between the AI agent and other tools or services.