An OS for AI workflows

Configure. Orchestrate. Observe. — AI pipelines that run the way you designed them.

Everything you need to orchestrate AI

A config-driven engine that puts the pieces together — whatever those pieces are.

Multi-stage pipelines with configurable stages, retry logic, and sequential execution. Define the workflow once, run it on every issue.

Route to any AI provider — Anthropic, OpenAI, Ollama — via config, not code changes. Fallback chains handle failures automatically.

A rules engine evaluates conditions between stages. Auto-approve, hold for human review, rework, or abort — per stage, per rule.

Define agent personalities, skills, and routing rules. Scope them globally or per-project with inheritance, forking, and overrides.

Stream every stage output live. Track tokens, costs, and success rates across all runs. Event-sourced from the ground up.

Docker Compose deployment. Your data stays on your infrastructure. AGPLv3 licensed — inspect, modify, and contribute.

Three steps from configuration to insight.

Define pipelines, personas, skills, and routing rules through the web UI or CLI. Everything is stored in the database — no config files to sync.

fluxaOS routes work to the right provider, materializes skills to the workspace, executes stages, and evaluates gates between them.

Watch runs stream in real-time. Track costs per provider, per model, per project. Measure outcomes and iterate on your configuration.

Get running in minutes with Docker Compose.

terminal

git clone https://github.com/fluxaOS/fluxaos.git
cd fluxaos
cp .env.example .env
docker compose up