HelixDB releases local open-source Helix Foundry stack

The team behind HelixDB has officially launched Helix Foundry, an Apache 2.0-licensed platform built as an open-source alternative to Palantir Foundry. Written entirely in TypeScript and designed to run on a single machine via Docker Compose, the system unifies disparate enterprise records into a shared graph ontology. It connects directly to PostgreSQL, MySQL, SaaS services like Stripe, WorkOS, and PostHog, as well as Amazon S3 object storage, standard file uploads, and generic HTTP APIs.

Under the hood, the architecture splits storage and query workloads across specialized engines. Entity relationships and ontology objects reside in HelixDB, while analytical SQL queries execute through DuckDB over versioned, immutable Parquet snapshots. For text-driven interaction, the project includes a local qwen3:4b model powered by Ollama. Developers who prefer cloud-based inference can optionally add API keys for Claude or OpenAI to assist with entity mapping and natural language SQL translation.

To keep sensitive enterprise data isolated, Helix Foundry enforces a local, single-user security framework bound exclusively to localhost without requiring an online account or hosted control plane. Database synchronization utilizes native change data capture for supported PostgreSQL and MySQL instances, while other connectors rely on scheduled snapshot updates. Security is further hardened by cutting off internet access and database credentials from the query engine, ensuring generated queries run solely against isolated Parquet files.

For data engineers and software developers, the tool offers a privacy-first sandbox to analyze unified business entities—such as customers, support tickets, and invoices—directly on local hardware. The environment includes an OpenAPI specification, a TypeScript SDK, and scoped API tokens, making it straightforward to build custom local integrations or programmatically evaluate organizational data stacks.