Filament by Galaxy

Filament: open-source data replication and CDC

Filament by Galaxy is an Apache 2.0 data movement engine written in Go. It moves data from databases, SaaS APIs, event streams and object storage into warehouses, lakes, databases and message buses, using full loads, incremental replication or change data capture.

How it works

• Verified, checkpointed batches: records move in bounded Arrow batches that are verified at the sink before a durable checkpoint advances, so failed runs resume instead of restarting.
• Write policies: replace, append, upsert or merge, validated against the read mode before any rows move.
• Connectors: PostgreSQL, MySQL, Kafka, NATS, GitHub, Linear, Notion, Slack, Stripe, HubSpot and more as sources; PostgreSQL, ClickHouse, Iceberg, BigQuery, S3, Snowflake, Redshift and others as sinks. Most SaaS sources are short YAML manifests.
• Agent-friendly: the repo ships an agent skill so a coding agent can research an API and write a connector.

Run it anywhere

Install the CLI with Homebrew or a one-line script, try the standalone binary or Docker image with its web UI, embed the engine in a Go service, or deploy on Kubernetes with Helm (server, control plane and per-run workers, PostgreSQL state, NATS JetStream events). Railway, Render and DigitalOcean templates, a REST API and Python/TypeScript SDKs are available.

Performance (vendor benchmark)

On a 298-million-row PostgreSQL-to-PostgreSQL test, Galaxy reports about 2.6M rows/s and 115 seconds wall time, and says Filament was 91x faster than Airbyte and used 15x less memory than dlt. The harness and raw results are public.

Pricing and company

Free and open source with no gated features; Galaxy offers deployment and migration help. Galaxy (Intergalactic Data Labs, Inc.) is based in New York and was founded by Mitch Bregman (CEO) and Leon Kozlowski (CTO). Filament is pre-1.0.