Ciaren
Open source · Local-first

Design pipelines visually. Export Python you actually own.

Drag and drop nodes, preview every transformation on real data, and export clean pandas or Polars code. Everything runs locally — no proprietary runtime, no lock-in.

View on GitHubRead the Docs
Ciaren visual workflow editor showing a pipeline that cleans data (Drop Nulls, Remove Duplicates) then trains, predicts, and evaluates a Random Forest model with the Train Regressor config panel open
Clean your data, then train, predict, and evaluate a model — all on the same canvas.

What Ciaren does

Go from idea to running pipeline in minutes

Drag file, SQL, transformation, and output nodes onto a canvas and connect them — no code required to start.

Know exactly what happened on every run

Every run shows per-node status and row counts, so a failure is easy to trace back to the exact step that caused it.

Catch problems before running the whole pipeline

Inspect the real output of any node on real data before you commit to running everything downstream.

Take your pipeline anywhere, without a rewrite

The Python Ciaren generates is the same code you'd hand-write — drop it into a script, a notebook, or a CI job.

Pick the right engine for the job

Switch between pandas and Polars per run — prototype fast, then scale without changing your workflow logic.

Run pipelines on your own schedule

Set up hourly, daily, weekly, or custom cron schedules and let Ciaren run without you watching.

Why Ciaren

Built different, on purpose

Most workflow tools trade transparency for convenience. Ciaren doesn't ask you to.

Other workflow tools
Ciaren
Requires a proprietary runtime
Exports plain, runnable Python
Cloud-first, hosted by default
Local-first — runs on your machine
Black-box execution
Readable, auditable code
Vendor lock-in
Open source (AGPL-3.0)
Heavy, warehouse-scale orchestration
Lightweight, prototyping-friendly workflows
How it works

From data to Python in five steps

Import data

Pull from CSV, SQL, or cloud storage to get started.

Build visually

Drag nodes onto the canvas and connect them.

Preview

Inspect the result of every step on real data.

Run locally

Execute the full pipeline on your own machine.

Export Python

Take clean pandas or Polars code with you.

From canvas to code

Visual workflow in, Python out

Every node you connect on the canvas maps directly to readable pandas or Polars code — no hidden runtime, no black box.

Load CSV
Filter rows
Group & aggregate
Export
pipeline.py
import polars as pl df = pl.read_csv("sales.csv") df = df.filter(pl.col("status") == "completed") result = (    df.group_by("region")    .agg(pl.col("amount").sum().alias("total_sales"))    .sort("total_sales", descending=True)) result.write_csv("sales_by_region.csv")

Built on tools you already trust

Python
FastAPI
React
Polars
pandas
scikit-learn
MLflow

Use cases

ETL

Join, reshape, and export data from files, SQL, and cloud storage — sketch the full pipeline before committing to production.

Data cleaning

Fill nulls, drop duplicates, and catch outliers with built-in transformation nodes, previewed on real rows as you go.

Machine learning

Train, validate, and compare models without leaving the visual workflow — every run tracked automatically in MLflow.

Feature engineering

Build and preview derived columns and transformations, and see the exact effect on your data before training.

Data validation

Catch schema drift, nulls, and outliers early with dedicated validation nodes — before bad data reaches production.

Prototyping

Sketch a pipeline in minutes, preview every step on real data, and decide what's actually worth productionizing.

Roadmap

  • Plugin SDK

    In progress

    Stable provider contracts with local and entry-point discovery already work; the lifecycle around them keeps improving.

  • Better ML nodes

    In progress

    Improving metrics, model comparison, lineage, and MLflow integration for the existing ML extension.

  • Official plugins

    Planned

    Production-grade plugin examples beyond the hello-world node, covering tests, packaging, and permissions.

  • Workflow templates

    Planned

    Starter flows for common classification, regression, clustering, and feature-engineering tasks.

  • Marketplace foundation

    Planned

    A lightweight index for discovering community nodes, connectors, templates, and exporters.

View full roadmap →

Developer friendly

GitHub repo

Browse the source, open issues, and track releases.

Documentation

Quickstart, examples, and the plugin guide.

Contributing guide

How to open issues, propose changes, and add plugins.

Security policy

How to report vulnerabilities responsibly.

Project status: Early alpha

Ciaren is under active development. APIs, workflow formats, generated code, and plugin interfaces may change before 1.0 — use it for experimentation, prototyping, and controlled internal workflows.

Open sourceAGPL-3.0AlphaPlugin SDKLocal-first

Ready to try Ciaren?

Open source, local-first, and in early alpha.

View on GitHubRead the Docs