Run AI pipelines
the way you designed them.
Stage-by-stage control, configurable quality gates between every stage, routing across any provider. Self-hosted and open source.
A config-driven engine
that puts the pieces together.
The only pipeline engine with configurable quality gates between every stage.
Auto-approve, hold for human review, trigger a rework pass, or abort — defined in config, not code. Devin won't let you. Continue can't write. fluxaOS does both.
Pipeline orchestration
Multi-stage pipelines with retries, sequential execution, and gates between stages. Define once, run everywhere.
Provider-agnostic routing
Anthropic, OpenAI, Ollama — route by config, not code. Fallback chains handle failures automatically.
Gate-controlled quality
Configurable quality gates between every stage. Auto-approve, hold for review, rework, or abort.
Configurable personas
Agent personalities, skills, routing rules. Scope globally or per-project — inherit, fork, override.
Real-time observability
Stream every stage live. Tokens, costs, success rates — event-sourced from the ground up.
Self-hosted, open source
Docker Compose deployment. Your data, your infrastructure. AGPLv3 — inspect, fork, contribute.
Three steps from config to insight.
Run it on
your own box.
Four commands. Docker Compose stack. Your data, your infrastructure.
$ git clone https://github.com/fluxaOS/fluxaos.git $ cd fluxaos $ cp .env.example .env $ docker compose up ✓ web http://localhost:3000 ✓ worker ready ✓ db postgres 16