Many software companies take feature requests in a Discourse forum, with a vote count on every topic. Ten of those forums were read on 4 October 2026: Cursor, Home Assistant, Brave, Bitwarden, n8n, Coda, Glide, Fibery, Logseq and Discourse's own forum, Meta. Every topic in their feature request categories was kept, 56,951 in all, with its votes, likes, replies, views, dates, state and tags.
Leaving out Meta, where votes are a recent habit, the median request has one vote and 38.7 percent have none. A request with no votes got no reply 43.2 percent of the time, against 0.8 percent for one with fifty or more. Bitwarden, which tags what it ships, marked 1.6 percent of its requests with no votes as released and 34.2 percent of those with fifty or more. On the nine forums still taking requests, the median open one was filed 1,364 days before the read.
host,category,topic,created,last_posted,closed,archived,pinned,votes,likes,replies,views,tags
community.bitwarden.com,pm-feature-requests,158,2018-03-08,2026-09-22,0,0,0,891,718,568,215676,app:desktop;in-progress;auto-type
community.home-assistant.io,feature-requests,102771,2019-03-01,2024-03-27,1,0,0,2169,428,325,53376,automation
forum.cursor.com,feature-requests,133736,2025-09-17,2026-06-25,1,0,0,614,2436,639,48418,
Download the file, one row per topic, compressed with gzip. It is free to copy, quote and use commercially. To cite it, link https://votefirst.app/en/data/discourse-feature-requests and give the read date, 4 October 2026. Each row names its forum and topic number, so https://forum.cursor.com/t/133736 opens the third row above.
How does voting work on a Discourse forum?
Through the Topic Voting plugin, switched on for chosen categories. A member votes with a counter at the top of the topic, once per topic. A forum can limit how many votes each member holds at a time, by trust level, and the plugin's default settings give a new member two and the most trusted ten. Bitwarden sets its own figure: 20 votes per member, once they have read enough of the forum to reach trust level 1.
When a topic is closed or archived, its votes are locked at the count they had, and each voter gets the vote back to spend again. The plugin also sends every voter a notification that their vote was released. Where limits are on, a vote on a Discourse forum is scarce in a way a thumbs up on GitHub is not, and a closed topic keeps the count it had on the day it closed.
How do the ten forums compare?
| Forum | Requests | Read from | Filed in the last year | Median votes | No votes | Ten or more | No reply | Closed |
|---|---|---|---|---|---|---|---|---|
| Home Assistant | 14,936 | March 2016 | none | 2 | 21.2% | 15.9% | 29.4% | 5.4% |
| Discourse Meta | 8,055 | February 2013 | 355 | 0 | 88.0% | 0.1% | 11.2% | 23.2% |
| Brave | 7,639 | May 2017 | 934 | 1 | 47.8% | 2.6% | 48.9% | 8.6% |
| Cursor | 6,697 | August 2023 | 2,982 | 0 | 54.8% | 4.4% | 40.6% | 40.4% |
| Bitwarden | 4,568 | March 2018 | 338 | 1 | 40.8% | 10.3% | 32.9% | 46.5% |
| n8n | 3,614 | September 2019 | 692 | 1 | 29.9% | 6.3% | 43.8% | 4.5% |
| Coda | 3,541 | February 2018 | 158 | 0 | 55.4% | 5.8% | 30.7% | 10.5% |
| Glide | 3,262 | September 2019 | 99 | 0 | 53.7% | 4.6% | 24.9% | 10.7% |
| Fibery | 2,275 | October 2018 | 446 | 0 | 56.4% | 2.4% | 23.0% | 18.9% |
| Logseq | 1,805 | December 2020 | 27 | 3 | 16.8% | 18.1% | 42.4% | 0.6% |
Home Assistant moved feature requests to GitHub and archived the category, so its last request was filed in June 2025 and most of its old ones were left open. Cursor's forum is the busiest by far: 2,982 requests in the twelve months to the read, about eight a day.
Meta is the odd one. 97 percent of its feature topics filed in 2014 have no votes, and the share falls year by year to 46 percent of those filed in 2026, while the median topic there has eight likes. With votes that sparse for most of its history, Meta is left out of the vote figures below.
Closed means closed to new replies and to voting. A forum records that a topic was closed, but not why: built, merged into another request, refused and closed on a timer all look the same unless the team adds a tag.
How many votes does a forum feature request get?
Usually one or none. Of the 48,337 requests on the nine forums other than Meta, 38.7 percent have no votes, 42.6 percent have one to four, and 8.9 percent have ten or more. 256 have a hundred or more, and ninety percent have nine or fewer.
That is more votes than a feature request draws on GitHub, where 72.1 percent of requests on the 1,000 most starred repositories have no thumbs up, and about the same backing as on a dedicated board. On 634 public Canny and Featurebase boards, the median request held two votes, one of them usually the asker's own. A Discourse request starts at zero.
Do votes bring replies?
They go together. A request nobody voted for was the likeliest to sit with no reply at all, and a request past fifty votes almost never did.
| Votes | Requests | No reply | Median replies | Closed |
|---|---|---|---|---|
| None | 18,725 | 43.2% | 1 | 22.8% |
| 1 to 4 | 20,572 | 38.3% | 1 | 12.2% |
| 5 to 9 | 4,736 | 18.9% | 3 | 7.8% |
| 10 to 49 | 3,672 | 6.9% | 7 | 9.1% |
| 50 or more | 632 | 0.8% | 36 | 19.6% |
A reply can come from anyone, so this does not say the team answered, and the counts cannot say whether votes drew the replies or the replies drew the votes. What they do say is that four in ten requests with four votes or fewer never got a second post.
Do more votes get a request built?
Only a forum that records what it built can say. Bitwarden tags a shipped request released, 190 so far, and because votes lock when a topic closes, each released request's count is the one it had when it shipped.
| Votes | Bitwarden requests | Tagged released |
|---|---|---|
| None | 1,865 | 1.6% |
| 1 to 4 | 1,877 | 2.7% |
| 5 to 9 | 355 | 6.8% |
| 10 to 49 | 354 | 13.0% |
| 50 or more | 117 | 34.2% |
Each step up in votes went with a larger share released. It was still a minority at the top: 68 of the 117 requests past fifty votes are open, among them the most voted on the forum, auto-type in the desktop app, filed in March 2018 with 891 votes and tagged in progress.
Three other forums show a process of their own. Logseq's guide asks members to have a developer tag a request with an effort estimate once it reaches 100 votes, and 8 of the 16 requests past 100 carry one. n8n moves finished requests to a separate category for done ones, which holds 487, against 3,614 still in the requests category. Discourse itself says in its guide that requests from its hosting customers come first.
Which requests have the most votes?
The most voted request on each forum, as it stood on 4 October 2026:
| Forum | Request | Filed | Votes | State |
|---|---|---|---|---|
| Home Assistant | Group automations in the frontend | March 2019 | 2,169 | Closed |
| Bitwarden | Auto-type in the desktop app | March 2018 | 891 | Open, in progress |
| Cursor | Student verification outside the US and without a .edu address | September 2025 | 614 | Closed |
| Brave | Support for the Cardano blockchain | January 2023 | 374 | Open |
| n8n | Choose credentials with an expression | April 2021 | 308 | Open |
| Logseq | Read EPUB ebooks | August 2021 | 285 | Open |
| Glide | Bring push notifications back | December 2024 | 241 | Open |
| Coda | Desktop apps | March 2021 | 136 | Open |
| Fibery | Relations that can point to more than one type | November 2019 | 51 | Open, marked approved |
| Discourse Meta | Show every like and reaction on a post | November 2025 | 30 | Closed, tagged completed |
Seven of the ten are open, and five of those were filed in 2021 or earlier.
How long do open requests wait?
A long time. On the nine forums still taking requests, 32,782 are open. The median was filed 1,364 days before the read, 83.5 percent more than a year before, and 58.7 percent more than three years before. On JetBrains' tracker, the median open feature request was filed 2,221 days before its read.
What does that mean for your own feature request board?
A forum puts the discussion next to the request, and on these ten the requests that drew votes also drew the conversation. What a forum does not keep on its own is the outcome. Closing a topic records only that it closed, and a closed request can be read as built only where the team tags it, as Bitwarden does.
On VoteFirst, Denied and Completed are columns on the board by default, so each request's outcome is on the board itself rather than in a tag. Voting is anonymous by default, with no forum account to create first, and when a release includes a request, every voter who was signed in when they voted is emailed that release. Picking feature voting software and adding a board to an open source project are each covered on their own page.
How was this read, and when does it change?
The ten are public Discourse forums whose feature request category runs topic voting. They are not a sample of all Discourse forums: each was picked because its category is open to read and shows a vote count on every topic. Each category's topic list was read through Discourse's public JSON, oldest first, on 4 October 2026, and every topic number in the file is distinct.
Pinned topics, usually the category's own guide, were left out. 487 topics in n8n's done category, where voting is off, and 72 on Fibery and Coda that show no vote count are in the file but not in the vote figures. Replies are posts after the first, which does not include the notice Discourse adds when a topic is closed. Votes are as shown on the read date, and on a closed topic that is the count it had when it closed. The file keeps no titles, posts or authors.
The next read is due by 1 January 2027.
14 days, a card at signup, then $9 or $39 a month.