Instead of a subscription

Self-hosted alternatives to GitHub

GitHub hosts repositories, issues, releases and Actions for most of open source, and for a great deal of private work. The forges below cover repositories, reviews, releases and CI runners for the work you would rather keep on your own machine.

Catalog

Recipes that cover it

Fully verified57,748

Development

Gitea

A lightweight Git service for repositories, issues, and collaboration

512 MB RAMamd64 / arm64GoMediumMITVersion 28.2.0
Fully verified5,411

Development

Forgejo

A self-hosted Git service for code, issues, and collaboration

512 MB RAMamd64 / arm64GoMediumGPL-3.0-or-laterVersion 16.0.5

Requirements

Side by side

Requirements and recipe status of the applications on this page
ApplicationMinimum RAMDisk estimateDatabaseInstallation difficultyBackupRecipe verificationLicense
Gitea512 MB RAM5 GBPostgreSQLMediumYes · Restore testedFully verifiedCheck is currentMIT
Forgejo512 MB RAM5 GBPostgreSQLMediumYesFully verifiedCheck is currentGPL-3.0-or-later

The bill

The subscription against a server

Both figures are editable, because only you know which of them you actually pay. The server is sized from the figure the recipe recommends — the same one the catalog calls a comfortable fit on your profile — and not from the cheapest row of a price list.

Gitea asks for 1 GB of RAM, 1 vCPU and 5 GB of disk — $6 a month on DigitalOcean Basic.

The difference

Per month
$2
Per year
$24
Over three years
$72

The subscription wins on price by $24 a year. What hosting it buys you here is the data, not the money.

Not counted: your own time, a domain, and a backup kept anywhere other than that same server. A machine you already run costs nothing extra — set the server to zero.

GitHub · Team, billed annually, price checked 2026-09-10Server prices: DigitalOcean Basic, checked 2026-09-10

Migration

What changes when you host it yourself

The data stops leaving your network, and so does the responsibility for it: backups, the restore you have actually tested, and the disk it all sits on.

Feature parity is never exact. Read what each recipe states about limitations and the smoke test before planning a migration.

The bill turns into work: updates, certificates and an eye on the release notes. The catalog shows how much of that each recipe asks for.

Categories

Other services in the same categories