Putting AI Agents to Work: A New Era of Open Source Connectivity with MCP
summary
Claim: MCP makes agents modular and interoperable by standardizing tool metadata, auth, and calls. The team builds a cloud-native stack with vLLM, Llama Stack, MCP servers for GitHub and Slack, and Kubernetes, then demos an agent that fetches GitHub issues and posts a summarized slack message. The demo loads 100+ tools, uses a read-only personal access token, and works end-to-end but shows noticeable latency and cost trade-offs; recommendations are narrow tool scoping, task specialization, and protocol-based isolation to reduce brittleness.
AI-written summary · about
chapters (9)
- 0:00AI agents introduction
- 1:06What are AI agents
- 6:12Failure modes in AI agents
- 12:15Building collaborative AI
- 14:42Model Context Protocol
- 18:00Automation use case
- 20:52Live MCP demo
- 25:21Implementation lessons
- 27:31Q&A
related talks
40mAgents in Action: Advancing Open Source AI Missions
30mComposable AI: Building Next-Gen AI Agents with MCP
46mMCP and A2A Revolutionizing Interconnection and Collaboration of AI Agents
32mFrom Monolith to AI Agent Modernizing Java Systems with MCP
16mOne AI Chat to Rule Them All: Connecting Enterprise Tools with MCP
29mBuilding Interoperable Agents: When A2A meets MCP
12mMCP and the Dawn of Agentic Intelligence
01h 23mA2A & MCP Workshop: Automating Business Processes with LLMs