Compose the Future: Building Agentic Applications, Made Simple
summary
Docker demonstrates concrete tooling: Compose gains a native Docker model runner keyword to pull OCI-packaged models, Compose wires models and tools to an MCP gateway backed by a 100+ entry MCP catalog, and a live sock‑store demo runs three cooperating agents (Reddit research, customer‑feedback, catalog) using GPT‑4 and a local Qwen‑3 model. Remote GPU offload launches larger models (demo shows switching to a 30B model), Cloud Run deploys the same Compose file serverlessly, and an automated agent fixes 144 vulnerabilities down to 0 and shrinks an image from 399 to 39.
AI-written summary · about
chapters (6)
- 0:00Why Docker Agents
- 8:30Docker Agent Features
- 17:32Sock Store Demo
- 31:20Docker Offload Reveal
- 33:02Serverless Cloud Run
- 43:02Key Benefits Recap
related talks
46mDocker Compose Your Dev Toolkit for AI & Cloud
25mBuilding AI Workflows: From Local Experiments To Serving Users
39mBuilding Agentic AI Apps with Docker MCP & Cagent
46mPowering Compose Multiplatform Apps with Koog & MCP
38mDesign and Implementation of Next‑Gen App Development Tools Using Compose Multiplatform × AI
29mFrom Copy-Paste to Composition: Building Agents Like Real Software
21mMaking Docker Actually Work
30mMind the Gap - A Developer's Roadmap To Building AI Agents