The self-hosted stack

Every one of these ships something you can run on your own hardware, with no vendor sitting between you and your data.

Self-hosting means running the software on hardware you control, which changes what you are buying. You stop renting a service and start operating one: the upgrades, the backups and the uptime become yours, and so does the data. That trade is worth making for some teams and not others. It is compelling when the data cannot legally leave your network, when per-seat pricing has outrun what a server costs, or when a vendor going out of business would take your work with it. It is a bad deal when nobody on the team wants to be on call for a Postgres instance. The tools here all publish something you can actually run, but they differ in how much they want you to. Some are open source with a hosted option as a convenience; others are hosted products with a self-hosted edition that lags behind and carries fewer features. Read which one you are getting before you commit, because the gap between those two is where the disappointment lives.

Supabase

Postgres with auth, storage and realtime attached, and no lock-in.

Developer toolsFreemium

Postgres with auth, storage and realtime built around it, and the whole thing starts from a compose file. The hosted tier is a convenience rather than the product, so moving onto your own hardware later is a deployment decision and not a migration.

PostHog

Product analytics, replays, flags and experiments in one open-source stack.

AnalyticsFreemium

Product analytics, session replay and feature flags in one deployment. Self-hosting matters more here than elsewhere on this list: what it collects is behavioural detail about your users, and keeping that on hardware you control is the whole difference.

Obsidian

Local markdown notes that link to each other and outlive any company.

ProductivityFreemium

Notes as plain markdown in a folder you already own. Less a server you host than software with no server at all, sync is optional and paid, and the files stay readable in any editor long after the app is gone.

Ollama

Run language models on your own machine with one command.

AIOpen source

Turns “run a model locally” into a single command, and exposes an OpenAI-compatible endpoint once it is up. That second part is why it anchors a stack. Everything else can point at it without any code changing.

n8n

Workflow automation you can self-host and extend with real code.

AutomationOpen source

Workflow automation you host yourself, which is rather the point. A hosted automation flow lives on someone else's servers with your API keys sitting inside it; these run on yours, and the credentials never leave.

Sentry

Errors with the stack trace, the release and the user who hit them.

Developer toolsFreemium

Error tracking, and a heavier install than most here: several services rather than one binary. Worth it because production stack traces routinely carry user data in their local variables, which is an awkward thing to hand to a third party.

Plausible

Web analytics with no cookies, no banner and a script under 1KB.

AnalyticsPaid

Web analytics without cookies, in one container. It answers the small number of questions most sites actually have, which is why the dashboard fits on a screen and why the self-hosted version is not a cut-down one.

Penpot

Open-source design and prototyping you can host yourself.

DesignOpen source

Design and prototyping in the browser, and the only tool of its kind you can run on your own infrastructure. Files are an open format rather than a proprietary blob, so the work survives the software.

Trigger.dev

Long-running background jobs in your own codebase, with retries that work.

AutomationFreemium

Background jobs and long-running workflows written in TypeScript, deployable to your own infrastructure. It fills the awkward gap between a cron entry that is too simple and a queue you would otherwise have to build and babysit.