Flynn PaaS

Open source · Self-hosted · Platform as a service

Ship software on infrastructure you own.

Flynn gives your developers the git push workflow of a hosted PaaS and gives your operators a cluster that runs on plain Ubuntu servers. Deploys, HTTPS, databases, scaling, and team access — with no vendor account in the loop.

Get started See the product Sign in

BSD-3 licensed. Source, releases, and docs on GitHub.

  • Open source, BSD-3-Clause
  • Runs on Ubuntu 24.04 LTS
  • Highly available on three nodes
  • Automatic Let's Encrypt HTTPS
  • CLI for macOS, Linux, and Windows

How it works

Install once. Push forever.

Flynn is a complete platform, not a kit of parts. Operators set it up; developers never touch a server again.

Install on your hosts

Run the installer on one or more Ubuntu 24.04 machines — bare metal, VMs, or a private cloud. Bootstrap the cluster, point DNS at it, and turn on Let's Encrypt.

Push code

Developers create an app and git push. Flynn detects the language and builds with buildpacks, or builds your Dockerfile, or takes an image you already built.

Route, scale, observe

Every app gets a URL and a certificate. Scale process types independently, attach databases, follow logs, and roll back when a release goes wrong.

What you get

The platform features teams actually use

Everything below ships in Flynn or one of its first-party plugins today.

Git, Docker, or GitHub deploys

Buildpacks for Go, Node.js, Python, Ruby, PHP, Java, and more. Dockerfile builds on the cluster. Pre-built images. Or connect a GitHub repository and deploy on push.

Releases with rollback

Every deploy and config change is a release. New releases start before old ones stop, and a failed release rolls back on its own.

Scale and runtimes

Scale web and worker processes independently. Give each process type a named CPU and memory runtime — small, medium, large, or your own.

Domains and HTTPS

Custom domains, path-based and TCP routes, sticky sessions, and Let's Encrypt certificates that renew themselves. Domains →

Datastores and volumes

Highly available Postgres is built in. MariaDB, MongoDB, Redis, Kafka, and ClickHouse install as plugins. Persistent volumes for processes that need disk. Data →

Team roles and isolation

Sign in through the dashboard, grant View, Deploy, Manage, or Admin per app, or build custom roles. Apps cannot reach each other's network. Trust →

Plugin catalog

Add what you need, when you need it

First-party plugins install with one operator command and update alongside the platform. Postgres is part of Flynn itself.

PostgreSQL 16

Built in. Primary plus synchronous and async replicas, with PostGIS, pgRouting, and TimescaleDB available.

Included

Dashboard

Web UI for apps, deploys, logs, releases, routes, team roles, alerts, and cluster settings.

App

Scheduler

Cron and interval jobs that run one-off commands inside your app's own release.

Jobs

OpenTelemetry

Forwards host metrics — CPU, memory, disk, load, job counts — to any OpenTelemetry collector.

Metrics

MariaDB 10.11

MySQL-compatible database with the same highly available primary and replica design as Postgres.

Datastore

MongoDB 7.0

Document database running as a replica set that promotes a new primary if one fails.

Datastore

Redis

Single-instance Redis with no failover. Data persists on a volume but is not in cluster backups — treat it as a cache.

Not HA

Kafka 3.9

Three-broker Apache Kafka in KRaft mode — no ZooKeeper — with TLS to clients by default.

Datastore

ClickHouse

Three-replica analytics database with ClickHouse Keeper and TLS, for event and reporting workloads.

Datastore

Discovery

Lets additional nodes join the cluster with a token served from the cluster itself.

Cluster

Homepage

This site. A public face for the platform on the www hostname and the root domain.

App

Enterprise

SSO, audit, cluster policy, and granular RBAC. Operators install it from the plugin catalog as enterprise.

App

Is it a fit?

What Flynn is — and what it is not

You will make a better decision with the full picture. Here it is.

Flynn is

  • A complete PaaS for web apps, workers, and the databases behind them.
  • Self-hosted on Ubuntu 24.04 servers you already run — a rack, VMs, or a private cloud.
  • Open source under BSD-3-Clause, with platform components that are themselves apps on the cluster.
  • Highly available on three or more nodes, including Postgres, MariaDB, and MongoDB failover.
  • A maintained community fork that keeps Flynn on modern Ubuntu with current language stacks and database versions.

Flynn is not

  • A hosted service. There is no Flynn cloud to sign up for. You run it.
  • A Kubernetes distribution. No manifests to write and no Kubernetes API — in either direction.
  • A sandbox for untrusted code. Apps are isolated from each other, but Flynn is built for your own team's software.
  • A big-data platform. The built-in databases suit development, staging, and small production — not very large datasets or heavy write volume.
  • Multi-region. A cluster wants its nodes on one low-latency network.

For operators

Built to be run by a small team

Flynn ships the operational tooling with the platform, so day two does not turn into a second project.

HA in three nodes

Start on one machine to evaluate. Move to three for production and survive the loss of a node.

Rolling updates

Update every host and system component from one command, with a backup-first workflow and pre-flight disk checks.

Backup and restore

One archive of app metadata, images, and Postgres, MariaDB, and MongoDB data restores onto new hardware. Redis, Kafka, and ClickHouse data is not included.

Isolation by default

Apps cannot reach each other or the platform's internals on the network. Containers run hardened and unprivileged.

Full operator documentation lives in the Flynn repository. What operating Flynn looks like →

Next step

Own the platform. Keep the speed.

Bring up a single node this afternoon, or read why teams choose to run their own PaaS.

Get started Why Flynn Sign in