The AI Native SDLC: Idea to Ticket to Working Code, Live

I've been putting together a workflow that uses Claude Code Plugins, a custom docs MCP server, and the Linear MCP to take an idea from rough sketch to work items that are actually ready to build, and then from tickets into working code. This talk is a live walkthrough of that.

I'll start with a brand new idea (nothing rehearsed) and take it through the whole flow in front of you. A PM-style prototype turning into scoped Linear tickets, Claude Code picking those tickets up, and doing the implementation work. Along the way I'll compare each step to how the same thing would play out in a more traditional flow, since that's where the value becomes obvious.

The stack:

• Claude Code Plugins for role-specific workflows, so the PM step, the engineering step, and the review step each have their own purpose-built experience
• A custom docs MCP server that gives every agent the same grounding in product context, architecture, and team conventions
• The Linear MCP so the output lands as real work items, not yet another tool your team has to remember to check

If you're a PM trying to hand engineering better-formed ideas, an engineer tired of untangling half-baked tickets, or a founder looking to cut down on handoffs between idea and code, I think you'll get something out of this one.

• Beginner
• Development
• AI/ML