Show HN: Dynamic code and feedback walkthroughs with your coding Agent in VSCode (www.intraview.ai)

A developer called Cy has released Intraview, a VS Code extension (compatible with Cursor-style forks) that creates dynamic, visual “tours” of code and agent activity to help developers maintain a systems-level mental model. Instead of just reading diffs or raw agent output, Intraview connects to your existing coding agent (examples listed: Claude, GPT5-codex) through a local MCP (middleware) to generate guided walkthroughs and feedback sessions—annotating PRs, agent-generated changes, or repos with contextual explanations and suggested edits. The project is presented as a raw demo rather than a polished product and emphasizes security-by-design: no external calls beyond anonymous telemetry sent to Azure for usage metrics. This matters because comprehension and alignment are growing bottlenecks as agentic coding scales: agents can write lots of code, but humans still need to understand, shape, and critique solutions. Intraview targets that human-in-the-loop gap by making agent reasoning and code structure explicit, enabling review, auditing, and directed feedback back to the agent. Technically, the local MCP architecture keeps sensitive code local while allowing multimodal tours and agent annotations; the extension could improve traceability, onboarding, and iterative design workflows for teams using LLM-based coding agents.