Back to all posts

How to Let Claude Analyze Your Feature Requests (2026)

SignalBoard

  • claude
  • ai agents
  • feature voting
  • duplicate detection
  • export
  • tutorial
  • product feedback

Published July 2026.

SignalBoard already holds the demand signal: titles, vote counts, statuses, and comments on your feedback board. You can export that board as JSON and paste it into Claude Desktop, Claude Code, or Cursor. Claude reads the snapshot, clusters themes, and helps you prioritize. No scrapers, no API glue, no manual spreadsheet cleanup.

Here is the workflow that matches what the product does today.

Step 1: Collect feedback on a public board

Customers use your board at /b/your-team/feature-requests (or a Pro custom domain). They verify email once with a magic link, then submit ideas, vote, and comment. Authors get an automatic vote on their own post, so new requests start at one vote.

Want a live example? Open the demo board. On Pro you can also make a board private; export still works from the dashboard for workspace members either way.

Step 2: Export JSON from the dashboard

Open the board in your dashboard. On the Posts page, click Export JSON, or go to Board settings → Export. Any workspace member can download it. The file is sorted by vote count (highest first) and includes only non-merged posts.

Each export looks like this:

{
  "board": "feature-requests",
  "exportedAt": "2026-07-08T12:00:00.000Z",
  "posts": [
    {
      "title": "Dark mode for the dashboard",
      "votes": 48,
      "status": "under_review",
      "comments": ["Need this for late-night on-call."]
    }
  ]
}

Fields in each post: title, votes, status, and comments (comment bodies only, no voter emails). Status values in the file are the internal ones: open, under_review, planned, in_progress, done, closed. On the public roadmap, done shows as Shipped.

Step 3: Paste into Claude and ask for a prioritization pass

Drop the JSON into Claude and use a prompt like this:

You are helping prioritize a SaaS feedback board.

Given this JSON of feature requests (title, votes, status, comments):
1. Cluster near-duplicates (do not invent posts that are not in the data)
2. Name the top 5 demand themes
3. Flag anything that looks like a bug, not a feature
4. Suggest what to ship next and why, using vote count + comment signal
5. Map suggestions onto these statuses when relevant: open, under_review, planned, in_progress, done, closed
   (done is shown as "Shipped" on the public roadmap)

Data:
<paste export here>

Useful follow-ups once Claude has the data:

  • "Which requests are the same idea under different titles?"
  • "Summarize what voters are saying in the top 10 posts."
  • "What should go on the roadmap this quarter given votes and comments?"
  • "Separate bugs from feature asks and rank both lists."

Claude does not write back to SignalBoard. You keep product judgment, then update statuses, merge duplicates, or publish a changelog in the dashboard.

Optional: clean the board with AI duplicate detection first

Before you export, let AI duplicate detection do its job on the public submit form. As someone types a title, SignalBoard embeds the draft with OpenAI, compares it to existing posts on that board, and shows up to three similar requests with a vote control. Nothing auto-merges. People can upvote an existing thread instead of creating another one.

That means the JSON you hand Claude already has fewer near-duplicate threads. You can still ask Claude to cluster leftovers. Duplicate detection is included on every plan with feature voting.

Close the loop in SignalBoard

After Claude suggests a cut, act in the product:

  • Move posts across public roadmap statuses (under review, planned, in progress, shipped).
  • Merge obvious duplicates from the dashboard so votes consolidate.
  • Publish a changelog entry when something ships. Voters and commenters can get email when a post moves to planned, in progress, or done.

What this is good for

Theme discovery. Group scattered titles into a few demand themes before roadmap planning.

Duplicate cleanup.Pair Claude's clustering with SignalBoard's submit-time suggestions and dashboard merge.

Roadmap drafts. Turn top-voted posts plus comments into a first-pass status cut you edit with your team.

Support and release copy. Pull comment themes into FAQ answers or changelog notes.

That's it

Export JSON. Paste into Claude. Bring the decisions back into SignalBoard. The board stays the source of truth for votes; Claude helps you think.

Start free, or try the live demo first. When you have a board of your own, use Export JSON on the Posts page and run the prompt above.

Collect demand signals your whole team can act on

Feature voting, public roadmaps, and changelogs with AI duplicate detection. Free tier, no credit card.

Free forever for your first board