Throw your user feedback in.
It comes back sorted.
Idea Radar is the idea board that sits next to your backlog. Support messages, feature requests, a note you scribbled after a call: drop them in raw. They group themselves into themes, duplicates merge, and each idea carries the number of people who actually asked for it.
Your backlog says what you decided to do. The Radar says what you heard and have not arbitrated yet.
Screenshots
Real captures of the app, added as this feature evolves.
“cannot find my old exports”
support chat“export is buried in a submenu”
public form“please add dark mode”
demoted ticketRaw user feedback in, an idea board out. Each theme carries the number of people who asked for it, not a score somebody invented.
Your users tell you things all day. A support message, one line in a chat, a note from a call, a feature request somebody filed eight months ago. Most of it is never read a second time.
Putting all of it in the backlog kills the backlog. A board where four lines out of five will never be built is not a plan, it is a graveyard with a sorting order. Putting it in the project memory is wrong too: memory holds what is settled, not what somebody wants. Leaving it in the chat means losing it the moment you scroll.
The Radar is the third place. Drop anything in, in whatever shape it arrived. It gets clustered into themes, deduplicated, and counted by the number of people behind it. When an idea is ripe you promote it to a backlog ticket, and everyone who ever asked for it gets told.
What an idea board actually buys you
User feedback arrives in bulk and comes back sorted. The themes are drawn from what people wrote, not from a taxonomy you maintain by hand and stop maintaining in March.
Every idea carries the number of people who asked for it. Product prioritization stops being a gut call and starts being a count you can show.
Duplicates are detected before they become two separate builds. Merging two ideas moves their signals, so the merge cannot drop a reporter.
Your backlog goes back to being what a backlog is for: what you are going to do, nothing else. Feature request triage happens on the Radar, not on the board your team works from.
Three things the data guarantees
An idea management tool is only worth the trust you can put in its records. Three properties are structural here, not policy.
Five ways to read the same corpus
One dataset, five questions. Switch lens and nothing moves in the data, only what you can see.
List
Every idea in one column, arrivals included, with a search field, a Status filter (Arrivals, To arbitrate, Ticket opened, Parked, Not planned) and a Theme filter, sorted by latest request, oldest request, latest activity, most requested or most voted. The lens the Radar opens on since version 1.193.
Which ideas carry this word, this status, this theme?
Method: filtered list
Themes
Ideas grouped into themes, themes sorted by heat, heat measured in people. This is the idea board you open in the morning to see what your users are talking about.
What are people talking about, and how loudly?
Method: affinity mapping
Impact / effort
A 2x2 matrix. Drag an idea and you are writing its impact and effort on the record itself. The top-left quadrant is what you should be doing right now, and it is populated by evidence rather than opinion.
What is strong and cheap, this week?
Method: impact effort matrix
Opportunity tree
Your objective at the top, themes below it, ideas under the themes. A theme attached to no objective shows up in an unattached column, which is the guardrail: it either gets connected to something you care about or it gets parked on purpose.
Does this serve my objective for the quarter?
Method: opportunity solution tree
Constellation
A semantic map built from embeddings. Ideas that mean the same thing sit next to each other, so duplicates are visible as clumps and gaps are visible as empty space. Merge candidates surface here before anybody builds them twice.
Where are my duplicates, and where are my holes?
Method: semantic clustering
The Radar and the backlog are one surface
An idea that is ripe becomes a ticket. A ticket you will not build this quarter goes back to being an idea. Both directions carry the people with them.
Bulk export of a project
11 signals, 24 votes
Ticket: bulk export
In progress, 11 followers
The link is visible from both ends: a promoted idea shows the card of its ticket, and a ticket born from an idea shows how many signals and votes it came with.
The link is visible from both ends: a promoted idea shows the card of its ticket, and a ticket born from an idea shows how many signals and votes it came with.
What happens to one piece of feedback
From the message somebody sent you to the ticket an agent ships, without anybody re-typing anything.
Something comes in
A support chat message, a submission on your public feedback board, a ticket you demoted off the backlog, a note you typed yourself, or something you dictated from your phone. Any of those is a valid way in.
It is stored as a signal
The raw verbatim, its author, its source and its date. Not a summary of it: the sentence itself. Whatever happens to the idea later, that record does not change.
It finds its idea and its theme
Classification attaches the signal to an existing idea or opens a new one, and the idea to a theme. If it looks like something you already have, it is flagged as a possible duplicate rather than merged behind your back.
The person hears back
The reporter gets a message in their thread: noted, here is the public link to the idea, you can follow it and vote on it. That single message is the difference between a user who reports again and one who never bothers.
You read the corpus
The list to find one idea by name, status or theme, themes to see what people talk about, the matrix to see what is cheap and strong, the tree to check it serves your quarter, the constellation to find your duplicates. Same data, five angles, no export.
An agent works the board with you
A Product Manager agent sits in the Radar panel with read access to the whole corpus. It reclusters, proposes merges, scores impact and effort and argues for the next ticket with the verbatims to back it. It proposes. Promoting an idea and answering a reporter stay your call.
Before and after you have a place for ideas
Nothing here is hypothetical. This is what happens to a backlog that doubles as a feature request board.
Without a Radar
- : Feature requests pile up on the board until nobody believes the priority order any more.
- : The same request arrives five times from five people and gets filed as five tickets.
- : The person who reported it never hears back, so they stop reporting.
- : Priorities come from whoever spoke to you last, or loudest.
- : Cleaning up means deleting, and deleting means losing the evidence that someone wanted it.
With the Radar
- : The backlog holds work that is actually going to happen, so its order means something again.
- : Five reports of the same thing become one idea with five signals and five names on it.
- : Every reporter gets a link, a status, a vote and a message when it moves.
- : Priority comes from a count of people, and you can open the list and read what each one wrote.
- : Cleaning up means moving to the Radar, and nothing is lost by moving.
The backlog and the Radar are not two tools. They are two states of the same request, and you can move between them in either direction.
What is an idea board, and how is it different from a backlog?
Is this a feature request board my users can post to?
How does automatic clustering of user feedback work?
How are duplicate feature requests detected?
Can users vote on ideas?
Does this give me a public roadmap?
What is the impact/effort matrix for?
What is an opportunity solution tree, and why is it here?
Can I send a backlog ticket back to the Radar?
Can I empty a whole backlog column into the Radar at once?
Does the AI product manager decide for me?
Can I use my own OpenAI key, and does it work on my phone?
You may also like
Public backlog
The feedback board your users already post to. The Radar is its idea tab, same address, same session.
Backlog task board
The kanban an idea is promoted into, where a ticket becomes an agent that writes the code.
Ticket scoping
Before you build a promoted idea, have a Product Manager agent frame it as a mockup of your real product.
Agent teams
Chain the agents that pick up a ticket once an idea has earned its place on the board.
Project memory
What is settled lives in memory. What is still just wanted lives in the Radar.
AgentsRoom MCP
The tool surface your agents use to read and reorganise the Radar from their own session.
Further reading
A Feedback Board for AI Agents: Let Your Users Write the Prompt
Feedback tools collect requests. None of them can build one. When the board your users write into is the same board your coding agents execute from, the rewriting step disappears.
Agents Write the Code Now. Here Is What the Developer's Job Became.
Writing code was one link in a six-link chain, and it is the link agents took. The other five got heavier. A walkthrough of the job that is left: hearing what to build, deciding, briefing, running, reviewing, shipping.
Give your users' ideas somewhere to live
Free to download, macOS, Windows and Linux, with the mobile app included.
Companion app: monitor your agents on the go
Bring your own: Claude, Codex, Antigravity CLI, or other AI provider.
Push bugs and requests straight to your public backlog.

