Skip to main content

Creating Agents

Build an AI agent from scratch in FlowStack AI Studio.

Step 1: Create a New Agent​

  1. Go to AI Studio in the dashboard
  2. Click + New Agent
  3. Enter a name and description for your agent
  4. Select the AI model (e.g., GPT-4o, Claude 3.5 Sonnet, Gemini Pro)

Step 2: Configure the System Prompt​

The system prompt defines your agent's personality, capabilities, and constraints.

Example system prompt for a customer support agent:

You are a helpful customer support assistant for FlowStack.

Your responsibilities:
- Answer questions about FlowStack features, pricing, and capabilities
- Help users troubleshoot common issues
- Escalate complex technical issues to the engineering team

Guidelines:
- Be friendly and professional
- Keep responses concise (under 200 words)
- If you don't know the answer, say so and suggest contacting support@onflowstack.com
- Never share internal system details or credentials

Tips for effective system prompts:

  • Be specific about the agent's role and boundaries
  • Include examples of good responses
  • Define what the agent should NOT do
  • Specify the tone and format of responses

Step 3: Add Tools​

Tools give your agent the ability to interact with external systems.

Available tool types:

  • HTTP Request — Call any REST API
  • Database Query — Query FlowStack Tables or external databases
  • Web Search — Search the internet for current information
  • File Operations — Read and write files
  • Integration Actions — Use any of FlowStack's 1,000+ integration actions

Example: Add a "Look Up Order" tool

{
  "name": "lookup_order",
  "description": "Look up an order by order ID and return its status, items, and shipping details",
  "parameters": {
    "type": "object",
    "properties": {
      "order_id": {
        "type": "string",
        "description": "The order ID (e.g., ORD-12345)"
      }
    },
    "required": ["order_id"]
  }
}

The agent will automatically decide when to use this tool based on the user's query.

Step 4: Configure Memory​

Set how your agent remembers previous interactions:

  • No memory — Each message is independent (stateless)
  • Conversation memory — Remember the current conversation (default, last N messages)
  • Summary memory — Maintain a running summary of the conversation
  • Vector memory — Store and retrieve relevant past interactions using embeddings

Memory window: Set how many messages to keep in context (e.g., last 10 messages).

Step 5: Test in Playground​

  1. Click Test Agent to open the chat playground
  2. Send messages and verify the agent responds correctly
  3. Test edge cases: unclear questions, out-of-scope requests, tool usage
  4. Iterate on the system prompt and tools until satisfied

Step 6: Deploy​

Deploy your agent as a REST API endpoint:

  1. Click Deploy
  2. Copy the API endpoint URL
  3. Use it in your applications or FlowStack workflows
curl -X POST https://app.onflowstack.com/api/v1/agents/your-agent-id/chat \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"message": "What is the status of order ORD-12345?"}'