Skip to main content

Create an AI Agent

Build your first AI agent in Agent Studio to automate tasks and workflows​

Creating an AI agent is the next step after setting up your Squid application. Agent Studio allows you to build powerful AI agents without writing code, connecting them to your data sources and systems.

Create an agent​

  1. Navigate to Agent Studio in the Squid Console Creating an agent - Step 1
  2. Click "Build your first agent" to start building your agent
  3. Provide an agent ID. It is a unique identifier (e.g., "banking-copilot", "support-assistant") that will be used to access and call it from code, or identify it later in the UI.

Note: This ID cannot be changed later and is used across the Squid SDK.

Creating an agent - Step 2

  1. Configure your agent - Customize the following settings:
  • LLM Model: Choose the underlying AI model that powers your agent
  • Agent Instructions: Define how your agent should respond to user input. Include response tone, behavioral guidelines, and step-by-step processes
  • Abilities: Add capabilities to enhance your agent (optional). Connect databases, upload documents, integrate APIs, or add custom functions

Creating an agent - Step 3

Test your agent​

Navigate to the Test Agent tab - Interact with your agent to validate behavior

  • Send test prompts and questions
  • View real-time responses
  • See behind-the-scenes processing
  • Iterate on instructions and abilities as needed

Agent Testing

Configure agent settings​

Access the Agent Settings tab to manage:

  • Access Control - Set agent to public or restrict with security rules
  • Audit Logging - Track usage, questions, and responses
  • Guardrails - Add compliance controls and safety measures
warning

Always restrict agent access before deploying to production. Public agents can be accessed by anyone without authentication.

Next steps​

After creating your agent, you can:

Add abilities - Connect databases, upload documents, or integrate APIs in the Abilities section

Implement security - Create security rules to control agent access

Build with SDKs - Extend your agent with custom code using the Squid SDK

Create workflows - Chain multiple agents together for complex automation

tip

Start simple with basic instructions, then gradually add abilities as you understand your agent's behavior. Test thoroughly in the dev environment before promoting to production.

Example agent configurations​

Customer Support Agent

  • Instructions: "Respond to customer questions about product features, pricing, and account issues. Always maintain a helpful, professional tone. Cite sources when providing information from documentation."
  • Abilities: Knowledge base (product docs, FAQs), Database connector (customer accounts)

Data Analytics Agent

  • Instructions: "Analyze sales data and create visualizations. Provide insights in clear, business-friendly language. Always include data sources and time periods in your responses."
  • Abilities: SQL database connector, Chart generation function

Document Processor

  • Instructions: "Extract key information from uploaded documents and summarize findings. Focus on dates, amounts, and action items. Format output as structured JSON."
  • Abilities: File upload (knowledge base), Custom extraction function