Capability Mesh : Architecture for the Agentic Era - Java Day Istanbul

Transcript

1. Capability Mesh Architecture for the Agentic Era Makan Sepehrifar April 2026
2. What is AI Agent? Agentic AI? Ref : Techcrunch “No one knows what the hell an AI agent is“
3. Who am I? Makan Sepehrifar • Senior Solutions Architect @ • Gen AI Adoption Specialist • Always a Developer
5. What is an Agent? “Autonomous, Goal-oriented computational system that perform certain tasks on behalf of a user, role or another program” Agent
6. Agent What is an AI Agent? An Agent that uses Artificial Intelligence (typically a Large Language Model) to achieve its goal. Large Language Model Reusable Prompt AI Agent
7. Agent Skills Ref : Skill.md Patterns by Bibek Poudel
8. Multi-Agent System Why we needed multiple Agents? • Parallel Execution and Maximum Output • Fault Tolerance and Resilience • Separation of Concerns and Specialization
9. (Multi-) Agentic System Break the Goal into Tasks Execute the Tasks Aggregate the Results
10. Agentic AI vs AI Agents Agentic AI Definition AI System that uses Agents (and not necessarily AI Agents) to achieve the Goal.
11. Agentic AI Break the Goal into Tasks Execute the Tasks Aggregate the Results Intent Detection Summarization Natural Language Natural Language Automated Agent Selection Shared Context in Natural Language AI Agent
12. AI Agent vs Agentic AI Plan a 5-day Amsterdam to Barcelona trip. Strict Budget : €1500 Plan a 5-day Amsterdam to Barcelona trip. Strict budget: €1500 (flights, lodging, activities). Generate a detailed itinerary focusing on historical sites and local food. Include the three cheapest flight options and three budget accommodations (hostels/budget hotels).
13. Why do we need multiple AI Agents? Ref : A small number of samples can poison LLMs of any size by Anthropic Homogeneity and Noise Reduction Context Purity Control, Monitoring and Observability Predictability and Integration
14. Agentic Architecture Emerging Challenge is How do these Agents (properly) communicate with each other? How to Scale? Is it possible to develop a manageable distributed agentic system that operates beyond the restrictions of local computing?
15. Same Problem, Different Solution?
16. 2010 Hype of Microservice Architecture How do Microservices communicate with each other? Is there a way to abstract the integration?
17. Service Mesh Architecture Encapsulate Integration Complexity in a Name
19. Capability Mesh Capability Mesh is a Multi-Agent Architecture where: • Agents are matched by their Capability (Intent). • Agent Execution is encapsulated in its Capability. • Unified Integration Protocol decouples Network and Agent Integration. • Execution Flow is audited to manage Agentic AI's Non-Deterministic Nature.
20. Architecture Elements Unified Communication Protocol Agent Discovery Controlled Execution
21. Unified Communication
22. Here come the Protocols A2A : Agent to Agent MCP : Model Context Protocol From AI Agent to AI Agent From AI Agent to any Agent (Tool)
23. "MCP is revolutionizing AI integration, just as REST transformed the internet." – Makan Sepehrifar , 2025
24. Agent Discovery
25. Again here are the Protocols A2A : Agent to Agent • Capability structured in Agent Cards ◦ https://<agent-host>/.well-known/agent.json • Community Project : a2a-registry MCP : Model Context Protocol • Standard MCP Registry by Anthropic ◦ Provide search and list of MCP Description • RPC 2.0 Built-In tools listing
26. Agent Discovery (with Agent Cards)
27. Agent Discovery (with MCP Registry)
28. Controlled Execution
29. Organizations use “Human Agent” as a risk mitigation mechanism. Human in the Loop Ref : Zombies in the Loop? by Sebastian Krügel Automation Bias (Over-trust) False Sense of Security Cognitive Fatigue Manipulation by AI Explanation
30. Human on the Loop Shift from in to on, where humans are seen as auditors not approvers. Policy as Code Self-Contained Agents Traceable Human Input LLM as a Judge • Deterministic Agents • Higher-Level Integration • Impact Validation
31. Capability Mesh
32. Capability Mesh (MCP Version)
33. Orchestrator
34. Orchestrator Responsible Entity in an Agentic (AI) framework aiming to achieve a bigger Goal. Key Components: • Workflow Management • Integration Capabilities • AI Integration Protocol Support • Agent Selection (Suggestion) • Intent Detection (Bigger Goal Definition) • Goal Verification Type of Orchestrators: • Workflow Managers • Orchestrators with Agentic Capability • AI Orchestrators
35. Orchestrator Semi-Deterministic Orchestrator: • Predefined Workflow • Predefined Task to Agent Mapping Non-Deterministic Residue: • Input Detection • Output Aggregation • AI Agent Behaviour
36. Orchestrator Non-Deterministic Orchestrator: • Predefined Agent Pool (?) Non-Deterministic Elements: • Input Detection • Agent Selection • Workflow Organization • Goal Verification • Output Aggregation • AI Agent Behaviour Planning Execution Verification Agent Pool
37. Orchestrator Hybrid Orchestrator: • Generative AI for Planning • Human-Agent Audit • Deterministic Orchestrator Execution Planning Audit Execution Workflow
38. Demo Time Show me the Code
39. Conclusion
40. Questions? https://www.linkedin.com/in/makan1869/ Makan Sepehrifar @makan1869