Developer Edition: Tutorial 3 - LLM Agents & MCP

In this video, Alistair advances the workflow from Tutorial 2 by introducing LLM Agents and the Model Context Protocol (MCP). You will learn how Dedoctive’s hybrid approach combines the structure of BPMN with the flexibility of AI agents (built on Google ADK), allowing them to autonomously select and utilize tools to complete tasks.

What You’ll Learn:
• Configuring LLM Agents: How to replace a simple script with a LLM Agent "Service Task", configure the model and define a response variable.
• MCP Server Integration: A step-by-step guide to adding an external MCP server. In this example, we add "Deep Wiki" (a public server) to give the agent access to GitHub repository information.
• Tool Selection: How to select specific tools (like ask_question or get_documentation) that the agent is allowed to use.
• Refining User Input: Updating the user form to accept specific queries which are automatically passed to the agent.
• Workflow Inspection: Running the workflow to watch the agent dynamically decide which tool to call, and inspecting the llm_response inside the Task Data to view the structured output.

Next Up: In the next tutorial, we will demonstrate how to call an MCP tool directly as a fixed part of the workflow, rather than having an agent decide when to use it.

Links:
• Github Repository: https://github.com/Dedoctive/DedoctiveDeveloperEdition
• Website link: https://dedoctive.ai/
#DedoctiveAI #DeveloperEdition #MCP #LLMAgents #GoogleADK #Gemini #workflowautomation

--- Tutorial Resources ---
Public MCP server URL: https://mcp.deepwiki.com/mcp

Видео Developer Edition: Tutorial 3 - LLM Agents & MCP канала dedoctive-ai