Show HN: Dynamic code and feedback walkthroughs with your coding Agent in VSCode

5 days ago (intraview.ai)

I've been programming since I'm 6 and I don't want to quit. Since Agents came into existence I've been increasingly building more of the random ideas.

BUT, like many I kept getting stuck and frustrated where I wanted to make changes with the Agent that I knew I could've made without it but I had *no clue* how things worked.

I created Intraview to help me build and maintain a mental model of what I was building (or had vibed) so I could use my knowledge to either fix it myself, or provide more directed instruction. It grew into something that's transformed my workflow in a pleasant way.

Intraview is a VS Code extension that allows you to create: - Dynamic code tours built by your existing Agent - Storage and sharing of tours (it's a file) - Batch Feedback/commenting inline in IDE in-tour and without (it's also a file)

Here's a video walkthrough for the show vs tell crowd where I jump in a random (Plotly JS) open source repo and build a tour to get started: https://www.youtube.com/watch?v=ROBvFlG6vtY

Talking tech design, this is very different than most because the whole App is cloudless. Not server less, there's no external APIs (outside basic usage telemetry).

  - basic TypeScript app, JS/CSS/HTML
  - Localhost MCP server inside VS Code (one per workspace open)

Three of the biggest challenges I faced was:

  - re-considering the user experience given there's no database 
  - trying to build a reasonable experience to manage MCP connection across so many different setups.
  - testing the many forks, Agents and themes because I wanted to make it look native (I'll probably reverse course here in future iterations)

What I'm curious about is, where do you see the value:

  - New project/developer onboarding 
  - PR reviews 
  - Keeping up with Agentic code 
  - Perf reviews (for EM), you could build a tour biggest contributions by a GitHub handle
  - Planning alignment and review with your Agent

You can see the extension page in VS Code with these custom links (Note: this redirects and requires permission to open VS Code, won't actually install, takes another click)

  - for VS Code: https://intraview.ai/install?app=vscode
  - for Cursor: https://intraview.ai/install?app=cursor

Once it's installed and you confirm MCP is connected to your local server, just ask your Agent:

  - Create an Intraview the onboarding for this app..
  - Let's use Intraview to gather my feedback on [whatever you created]. Break down steps such that I can provide good granular feedback.

Looking forward to your feedback and discussion.

And because this is HN. A relevant quotable from PG.

  “Your code is your understanding 
  of the problem you’re exploring. 
  So it’s only when you have your code 
  in your head that you really understand
   the problem.” 
  
  — Paul Graham

12 comments

cyrusradfar

Reply
mayahisali  4 days ago

Hi, Good job, I have two questions:

- The cloudless architecture is intriguing. How do you handle tour synchronization when multiple devs are working on the same codebase? - How do you handle tour updates when the underlying code changes? Auto-invalidation or manual refresh?

Thanks

  • cyrusradfar  4 days ago

    Hi! Thank you :) Great question and it's one that's full of assumptions so tell me where you think I'm wrong.

    Tours are stored as a flat file (JSON). When you open a tour, you open it with the Agent and on, re-opening, it goes through re-validation (as is done during the tour). In the case where lines shift in files because the file includes start/end ranges and line text (verbatim), this will cause the tour to be invalidated and the Agent will need to rebuild it. There are a lot of "dead code" cases where this won't catch a tour update, today.

    The tour file contains a ton of meta information, as well, so each time the Agent opens it -- it could rebuild the tour with the same intents. The original user request, synthesized request, success criteria, concepts that are going to be taught.

    This is a first pass solution for now. If there's more interest, there's a much more interesting tech solution with managing pointers to references and invalidating the tour on a broader range of criteria.

    Part of what I need to learn is:

      - A. Are tours something that folks use as ongoing "interactive" documentation, e.g. a set of onboarding tours
  - B. Are tours ephemeral that folks use to hash out how to approach a feature, bug or review a PR.

    At the moment, it's more optimized for B as that's how I've been using it but A is a very cool use case because one could imagine opening a new repo and having N tours that you can just start up. I was considering reaching out to open source projects to create tour-files and request a PR after they confirm interest (not spam them) to test A.

    Your thoughts are welcome and appreciated. Have a lovely Friday.

mock-possum  4 days ago

Huh wow usually I’m pretty skeptical about stuff like this but your video demo looks pretty neat, I’m gonna try it on my our codebase at work today and see how it goes! We’ve moved fast and broken stuff lately and struggled a bit to come up with coherent contributing guidelines, in addition to onboarding new devs and guiding LLM codegen - it’d be cool if a tool like this could help elucidate the key things you need to know to work within our bespoke framework.

  • cyrusradfar  3 days ago

    Ah, shucks! Thank you. Means a lot to me to have folks giving it a go. I recognize how much our attention is strained these days.

    I'm trying to figure out whether folks really want this as more of a documentation tool for onboarding, or for "on-the-fly" transient digging: bugs/root cause, PR Review support, co-designing, or code style review across a section. The inline feedback is what's been clutch and sticky for me both within tours, and outside.

    If you have feedback and install it or have issues you can email me at support at v1 d0t co --

    If it's more of an idea, feature request, or potential bug you can ask your agent (once MCP is connected) to "Send feedback to intraview team" or something similar to be clear you don't want to add feedback for the Agent (which it can also do) and it'll send us a Slack message.

joshribakoff  3 days ago

I see value in backfilling missing / obtuse documentation, i also see potentially negative value if used instead of improving or reading existing docs. Ideally Plotly would have a developer guide that is higher quality than what an LLM can dervive.

  • cyrusradfar  3 days ago

    That's really insightful. I think the goal is faster comprehension for the developer. Whether it's a PR in an area of code you haven't seen in a while and you want a refresher, a bug that you have no clue where to start, or you're onboarding -- in all cases, you can't start without a mental model of what's going one.

    I'm solving may be for folks like me who would prefer to have the guide IN CODE. It's a new experience to me as I've never used an in IDE code tour.

    I think what I'm experiencing is it's way more helpful for me to have the guide navigate through code so I can explore a little at each step and get a lay of the land.

    How people learn is personal, and what I hoped for when I was building this was more of an experience where I'm walking through with a teammate.

    I considered trying to have a voice AI read or explain each step. I also considered allowing another person to drive your IDE through mirroring what they're looking at. Both were cut at the idea phase because it felt it was feature bloat on a concept that I didn't know anyone (but me) really wants

    Open to any more ideas or feedback! Thank you so much for dropping in

jay_devs  4 days ago

Gave this a spin and this is really cool! Wish I could provide more in-depth feedback to help improve it but will certainly be keeping an eye on it! Nicely done!

  • cyrusradfar  2 days ago

    Hi @jay_devs! Thank you.

    > Wish I could provide more in-depth feedback to help improve it

    I've got updates on this thread here, email me at support at v1 d0t co, or you can tell your agent to write up/synthesize your thoughts and submit feedback to intraview (assuming you have the extension installed).

    If you're saying you don't have time -- totally get it and no presh! You made a builders day by giving it a go.

rcarmo  4 days ago

Hmmm. Why Claude Code and not Copilot?

  • cyrusradfar  4 days ago

    Hi, I just didn't test with Co-pilot but it should work as I think it's using some flavor of GPT4 or 5, both of which were able to use the new tools. I've tested more with GPT5 through cursor.

    Given we're all here to learn, I'll share more details on the testing matrix:

    IDE: VSCODE, Cursor, Windsurf, Cline, Roo

    Agents: Claude Opus/Sonnet/Haiku, OpenAI GPT-5/GPT-5-codex-[low|med|hi], etc...

    Theme: Then when you hit UI, if you use themes there's at four that are very common including default light/dark modes.