Self-hosting a feature request board

Self-hosting a feature request board means running the software that collects your customers' bug reports and feature requests on infrastructure you control, rather than paying a vendor to run it for you. The appeal is obvious: no per-user pricing, no vendor deciding your data's future, and a board that lives on your own domain. The catch is that "free" describes the licence, not the operation.

This guide is the honest version. It covers what the options actually are, what they cost once you include the parts nobody quotes, what genuinely breaks, and the case for not doing it at all.

What does self-hosting a feature request board actually involve?

Most open-source boards in this category are conventional web applications. That means a server process, a PostgreSQL database, a mail provider for the notification emails, TLS certificates, backups, and a plan for applying security patches. Fider, the best-known option, documents Docker and PostgreSQL 12 or later as requirements, plus your own SMTP or Amazon SES.

That list is the whole argument. Every item on it is a thing that can break at three in the morning, and none of them is the feature request board.

The exception is a board built as a static site. If the posts are files in a repository and the pages are generated at build time, there is no database to run, no container to keep alive, and no mail server to configure. That is a genuinely different operational profile, and it is the one Docket uses.

What does it really cost per month?

Here is the most honest number published by anyone in this category, and it comes from a competitor arguing against self-hosting. UserJot's comparison page costs a self-hosted Fider deployment at $70 to $125 a month in infrastructure before any engineering time: roughly $20 for Postgres, $20 to $40 for the host, $10 to $35 for transactional email, and $20 to $30 for monitoring and TLS. They put the ongoing effort at five hours a month, which they note is what teams quote once they have actually deployed it.

Quackback, an open-source project selling the opposite conclusion, puts it lower: $10 to $20 a month on a VPS for a small team, rising to $50 to $100 for a larger one.

Both are probably right, because they are describing different things. A managed Postgres instance with monitoring costs what UserJot says. A single small VPS running everything costs what Quackback says, and carries the risk that the single VPS is also your backup strategy.

Now put a subscription next to it. Checked against each vendor's own pricing page on 20 August 2026, and Canny.io's ladder against its help centre on 22 August 2026:

At $70 to $125 a month, self-hosting a database-backed board is not obviously cheaper than a $22 to $29 subscription. That is the part the open-source enthusiasm usually skips, and it is why this guide starts with it.

The maths only changes decisively in two situations: when the hosted tool prices by something that grows, or when the self-hosted thing has no server to pay for.

Which pricing models grow underneath you?

Canny.io prices by tracked user, and its own pricing page is clear about what that means: anyone who posts, votes or comments. Those users accumulate permanently. The count only goes up.

The consequence is that a three-year bill is not three times year one. Billed yearly, the entry tier is $79 a month at 100 tracked users and $1,079 a month at 5,000. Billed monthly, the same two points are $99 and $1,349. Canny.io's page also states that exceeding your limit means Canny.io "automatically upgrade[s] you to the next tier" with a prorated charge unless you set a spend limit, and that once that spend limit is reached Canny.io "will stop tracking feedback from new users".

Featurebase prices per seat, at $29 per seat per month billed yearly. Three colleagues is $1,044 a year before you add anything.

If your board succeeds, both of those bills grow because it succeeded. A self-hosted board does not care how many people vote.

What are the actual open-source options?

Checked 21 August 2026.

Four things worth knowing before you pick one.

Astuto is gone. The repository was archived read-only on 8 February 2026, the maintainer posted an end-of-life notice pointing people at Fider, and the hosted service ran for under a year before shutting. Several listicles still list it as a live option.

Fider is a company now. getfider.com redirects to fider.io, which is a commercial hosted service run by Northern App Labs Ltd, a UK company, with a pricing page. The code is still AGPL and still maintained. But their own documentation says "the easiest way to get started with Fider is to use our Cloud Instance", and self-hosting appears on their marketing site in a single footer link. The best-known self-hosted option markets hosting.

LogChimp is open core, and the gate is the product. Boards, Roadmaps and Comments are Pro-only at $24 a month billed yearly, on the build you host yourself. What free self-hosting gives you is posts, customers, one admin role and a custom domain.

Nobody here has a changelog. Fider has roadmap and voting and private boards, and no changelog at all. If you want the board, the roadmap and the changelog, you are running more than one thing.

Each of these is compared in more detail, with its licence, star count and current state, in the open source options compared.

What actually breaks?

Patching, mostly, and it is not hypothetical. Fider shipped v0.35.0 fixing a critical authentication vulnerability, v0.36.0 fixing SSRF and cross-site scripting, and v0.36.1 with further security work, all within three months. Noticing each of those and applying it is yours. A hosted vendor does it without telling you.

Email deliverability is the second one. A feature request board that cannot send "somebody replied to your post" is a board people stop returning to, and getting SPF, DKIM and DMARC right on your own domain is a genuine afternoon. There is a whole page on that: SMTP for self-hosted tools.

Backups are the third, and the failure mode is specific: most people have a backup and have never tested a restore. A database dump you cannot restore is not a backup.

When is a hosted tool the better answer?

There is a real, public, well-argued case against self-hosting, and it is worth reading before you commit. On 29 October 2025, someone on Second Life's public feedback board filed a request titled "Replace Canny with Selfhosted Alternative", proposing LogChimp on cost and data-ownership grounds. Maestro Linden declined it officially. The comments carry the honest counter-argument: Linden Lab would become entirely responsible for patching and stability, and they had already moved to Canny.io from self-hosting, to reduce exactly that burden.

That is the whole trade in one thread. Somebody else running it, patching it, backing it up and answering for the uptime is worth real money to a team with nobody to own it.

Choose hosted if you have no one who wants to be on call for a support tool, if your tracked-user count will stay small enough that per-user pricing never bites, or if you need SOC 2 evidence you do not want to produce yourself.

The three questions that decide it

1. Does the hosted price grow with your success? If it is per tracked user or per seat, model three years, not one month. 2. Is there a server in your self-hosted option? If yes, budget $70 to $125 a month and five hours. If it is static files in a repository you already have, most of that ledger disappears. 3. Who gets paged? If the answer is nobody, pay somebody.

Docket exists because of the second question. It is a static site plus two small endpoints, so there is no Postgres to run, no Docker, no mail server, and the free tier of any static host covers it. The posts are Markdown files in your own repository, which is also the backup.

That does not make hosted tools wrong. It makes the cost of self-hosting a question about architecture rather than a question about principles.

Where to go next

This page is the overview. Each part of it has a page of its own.

• What it costs to run. What a self-hosted support centre costs to run itemises the monthly bill.
• Deciding at all. Self-hosted or hosted, an honest decision table runs your own situation through it, and what actually breaks when you self-host takes the best public argument against self-hosting seriously before answering it.
• The architecture question. Static site or server works out which jobs actually need something running, and uptime, static against hosted compares the failure modes rather than quoting percentages nobody can verify.
• Choosing where it runs. Netlify, Vercel or Cloudflare Pages, including the commercial-use terms that catch people out, running it on a cheap VPS for the bill nobody itemises, self-hosting on Hetzner, Docker, Coolify, Railway, or nothing at all, and self-hosting on a Synology NAS.
• Getting it reachable. Custom domains and subdomains, and Cloudflare Tunnel when the thing is behind a home connection with no public address.
• Keeping it working. SMTP for self-hosted tools so the notifications arrive, backing up a feature request board so the restore works, the security headers a public feedback page needs, and spam on a public feedback form.
• The data you are now responsible for. GDPR data residency when you host it yourself, storing customer email addresses in a git repo, and git history is permanent, which is the sharp edge of keeping feedback in a repository.
• Access control. Putting a feature request board behind your SSO, and why most public boards should not be.
• What free and open source actually covers. Open core, and what is actually free, which is the licence question underneath every option above.
• The neighbouring things you may also be self-hosting. Self-hosted status page tools, where one option runs with no server at all, and open source knowledge base software, where two of the best-known options are not open source in the way people assume.
• Letting an agent run it. The support inbox as a git repository is the shape, an AGENTS.md for your support centre is the instructions, letting an agent answer without letting it invent is the part that matters most, an MCP server, or the files already in your repo is why this needs less plumbing than it looks, and what it costs to run support through an AI provider is the bill.

Frequently asked questions