Agentic Engineering & Tooling Agentic Engineering & Tooling Deep Dive 180min 216 Agent Loops, Decoded: Fundamentals to Agentic Patterns in Spring AI

This talk introduces Spring AI agents as context orchestrators, starting with ChatClient, memory, Advisors, RAG, Tool Calling, and MCP. It then builds advanced agent patterns: tool discovery, evaluation loops, explainable agents, planning, user questioning, subagents, and agent-to-agent integration, previewing Spring AI 2.1’s typed multi-step agent abstraction.
Video: https://youtu.be/Q91obeWeWNQ

Transcript

2. Building Blocks - System Two
3. Two Ways to Manage Context Hard problems need rounds of both. When the model drives the rounds, that is the Agentic Loop.
4. Building Blocks - SystemOne
5. The 4 Primitives
6. Fundamentals
7. Relentless Context Curation
8. Unstructured → Structured Output
9. Stateless → Session Memory Advisor
10. Spring AI Advisor
11. Spring AI Advisor - Recursive Loop through the advisor chain multiple times by calling chain.copy(this).nextCall(...) instead of chain.nextCall(...). Each iteration goes through the full downstream chain, maintaining proper observability.
12. Spring AI Advisor - Chaining First In Last Out (FILO)
13. Knowledge Cutoff → Context Stuffing
14. Knowledge Cutoff → Semantic Search (RAG) (runtime) Retrieving (offline) Loading
15. Knowledge Cutoff → RAG Advisor
16. Knowledge Cutoff → Modular RAG + Jev Not needed for Agentic RAG
17. Knowledge Cutoff → Modular RAG + Jev - Advisor
18. Hallucinations → Guardrails
19. Hallucinations → Guardrails: Structured Output Validation
20. Hallucinations → Guardrails: Jev Guardrail
21. Hallucinations → Guardrails: Jev Guardrail
22. Autonomous LLM → Tool Calling
23. Autonomous LLM → Tool Calling Advisor
24. Chaining Advisors
25. Progressive (Tool) Disclosure → Tool Search Tool
26. Spring AI : API & Extensions
27. Foundation Stack : ChatClient, ChatModel, Advisors, Tools, Memory, RAG, Guardrails, … Models VectorStore Multimodality Tool Calling Struct. Out. Prompt Memory Chat Model Advisors Chat Client Embedding Model Image Model Audio Model ETL Modular RAG Moderation Model Model Context Protocol (MCP) Observability …
28. Agentic Systems & Patterns
29. Brain LLM Reasoning, Planning Memory Body Tools Interacts with its environment Agentic Loop Leverages an AI models to interact with its environment in order to solve a user-defined task Relentless loop of assembling context, observing the result, and then re-assembling a context for the next step
30. Agentic Workflows Chain Workflow Orchestration - Worker Workflow Break complex tasks down into simpler, more manageable steps Central LLM orchestrates task decomposition. Specialized workers handle specific subtasks Routing Workflow Complex tasks with different input types, handled by specialized processes. An LLM analyzes the input content and routes it to the specialized handler. Evaluator - Optimizer Workflow Dual-LLM process - one LLM generates responses while another provides evaluation and feedback in an iterative loop Parallelization Workflow Work simultaneously on tasks and aggregate outputs Building effective agents: https://www.anthropic.com/engineering/building-effective-agents Building Effective Agents with Spring AI (Part 1): https://spring.io/blog/2025/01/21/spring-ai-agentic-patterns GitHub Repo: https://github.com/spring-projects/spring-ai-examples/tree/main/agentic-patterns
31. LLM-as-a-Judge (Evaluator - Optimizer)
32. LLM-as-a-Judge (Evaluator - Optimizer) - Advisor
33. Model-as-a-Judge (Evaluator - Optimizer)
34. Model-as-a-Judge (Evaluator - Optimizer)
35. Acausality → Explainable Agents
36. A Brain Needs Hands → Core Agent Tools Group Tools Role Files Read, Write, Edit, Glob, Grep, ListDirectory See and change the workspace Shell Bash, BashOutput, KillShell Run anything, including skill scripts Web WebSearch (Brave), SmartWebFetch Reach current information Environment AgentEnvironment Tell the model where it is: OS, directory, git status
37. Progressive (Context) Disclosure → Agent Skills https://agentskills.io/home#adoption
38. Progressive (Context) Disclosure → Agent Skills
39. Agent SkillsJars https://www.skillsjars.com/docs
40. Lost In the Middle → Todo Write
41. Over-opinionated → Ask User Question
42. Context Segregation → Sub-Agents
43. Sub-Agent: A tool whose body is another loop
44. Sub-Agent: Another markdown file
45. Sub-Agent: Demo
46. Durable Memory → Auto-Memory
47. Durable Memory → Auto-Memory Advisor
48. AI Protocols
49. AI Protocols - MCP, ACP, A2A
50. Model Context Protocol (MCP)
51. Non-MCP AI Architecture AI App 1 Database AI App 2 Document Repo AI App 3 External API Service Generative AI Generative AI Generative AI AI App Developer Responsibility M x N integrations
52. MCP AI Architecture Model Context Protocol AI App 1 Generative AI Database MCP Clients Document Repo AI App 2 Generative AI AI App 3 External API Service MCP Servers Generative AI AI App Developer Responsibility Service Provider Responsibility M + N integrations
53. Client/Server Architecture (MCP Host) AI Application MCP Client #1 … Generative AI MCP MCP Server #1 MCP Server #2 MCP Client #M … Host MCP Server #N AI App Developer Responsibility API API API External Services #1 External Services #2 External Services #N Service Provider Responsibility
54. MCP Client & Server Capabilities MCP Client has MCP Server MCP Transport has Features Features Roots Tools Sampling Resources Prompts Elicitation Completion Logging Shared Features Progress Ping Cancellation
55. MCP Pluggable Transports In Host Process MCP Client MCP Server STDIO Stateful Client (Host) Process Stateless Streamable HTTP Server Process MCP Server MCP Client HTTP/SSE
56. MCP Inspector https://github.com/tzolov/voxxeddays2026-demo/tree/main/17-mcp
57. MCP Apps https://spring.io/blog/2026/03/18/mcp-apps
58. Agent To Agent Protocol (A2A)
59. Agent2Agent (A2A) https://spring.io/blog/2026/01/29/spring-ai-agentic-patterns-a2a-integration
60. Agent2Agent (A2A) Subtasks https://github.com/tzolov/voxxeddays2026-demo/tree/main/18-a2a
61. Agent Client Protocol (ACP)
62. Agent Client Protocol (ACP) https://github.com/agentclientprotocol/java-sdk
63. Wrap Up
64. Spring AI Observability
65. References - Documentation & Projects Spring AI: docs.spring.io/spring-ai/reference/2.0 Spring AI Agent Utils: https://github.com/spring-ai-community/spring-ai-agent-utils Spring AI Tool Search Tools: github.com/spring-ai-community/spring-ai-tool-search-tool Spring AI A2A: github.com/spring-ai-community/spring-ai-a2a MCP Java SDK: modelcontextprotocol.io/sdk/java/mcp-overview A2A Java SDK: github.com/a2aproject/a2a-java MCP Security: github.com/spring-ai-community/mcp-security Agent Skills: agentskills.io/specification
66. References - Agentic Patterns Recursive Advisors: spring.io/blog/2025/11/04/spring-ai-recursive-advisors LLM-as-a-Judge: spring.io/blog/2025/11/10/spring-ai-llm-as-judge-blog-post Tool Search Tool: spring.io/blog/2025/12/11/spring-ai-tool-search-tools-tzolov Explainable Agents: spring.io/blog/2025/12/23/spring-ai-tool-argument-augmenter-tzolov Part 1 - Agent Skills: spring.io/blog/2026/01/13/spring-ai-generic-agent-skills Part 2 - AskUser: spring.io/blog/2026/01/16/spring-ai-ask-user-question-tool Part 3 - TodoWrite: spring.io/blog/2026/01/20/spring-ai-agentic-patterns-3-todowrite Part 4 - Subagents: spring.io/blog/2026/01/27/spring-ai-agentic-patterns-4-task-subagents Part 5 - A2A: spring.io/blog/2026/01/29/spring-ai-agentic-patterns-a2a-integration
67. References - MCP, Demos & Where to Start MCP Boot Starters: spring.io/blog/2025/09/16/spring-ai-mcp-intro-blog Dynamic Tool Updates: spring.io/blog/2025/05/04/spring-ai-dynamic-tool-updates-with-mcp MCP Apps (Rich UIs): spring.io/blog/2026/03/18/mcp-apps MCP Server Security: spring.io/blog/2025/09/30/spring-ai-mcp-server-security Voxxed Days Amsterdam 2026 Samples: https://github.com/tzolov/voxxeddays2026-demo Agent Utils Examples: github.com/spring-ai-community/spring-ai-agent-utils/tree/main/examples Spring AI Examples: https://github.com/spring-projects/spring-ai-examples Flight Booking + MCP: github.com/tzolov/playground-flight-booking Flight Booking MCP Server: github.com/tzolov/flight-booking-mcp
68. Thank You