Helix Foundry Lets Developers Query All Their Business Data on a Laptop
Helix Foundry lets you wire databases, SaaS tools, and files into a single company ontology that runs entirely on your laptop.
- HelixDB released Helix Foundry, an Apache 2.0 open source alternative to Palantir Foundry.
- Connects databases, Stripe, WorkOS, PostHog, S3, and files into one company ontology stored as a graph.
- Runs fully local via Docker Compose with HelixDB for the ontology and DuckDB for analytics.
- Ships with a local qwen3:4b model via Ollama, or optional Claude and OpenAI keys for AI queries.
- Includes change data capture, Parquet snapshots, OpenAPI, a TypeScript SDK, and scoped API tokens.
- Single-user security model: no accounts, bound to localhost, not meant to be exposed over the network.
Helix Foundry runs an ontology-driven data stack on a laptop
The team behind HelixDB has released Helix Foundry, an open source platform for connecting databases and SaaS tools, modeling their records as a shared graph, and querying the result with SQL or natural language.
Released under the Apache 2.0 license and written in TypeScript, the project runs as a Docker Compose stack on a single machine. It requires no Helix account or hosted control plane. Imported data and credentials stay in the local deployment, while the bundled Ollama model keeps inference on the host. Choosing Claude or OpenAI sends the prompts and context used for those requests to the selected provider.
A graph over local snapshots
In Foundry, an ontology is a common model for business entities such as customers, invoices, users, and support tickets, along with the relationships between them. The software examines connected sources, proposes entities and identity matches, and lets the user review the resulting model.
| Component | Role |
|---|---|
| HelixDB | Stores ontology objects and relationships as a graph. |
| Parquet snapshots | Preserve imported source rows in immutable, versioned files. |
| DuckDB | Transforms snapshots and executes analytical SQL in isolated processes. |
| Language model | Proposes the ontology and translates natural-language questions into SQL. |
| Web interface and API | Expose metrics, query results, workspace controls, and developer access. |
The query executor has no internet access or source credentials. Generated SQL runs only against the imported snapshots, limiting what a faulty or adversarial query can reach.
Sources in, SQL out
Built-in ingestion spans relational databases, selected SaaS products, object storage, file uploads, and a generic API. Sync behavior depends on the source: PostgreSQL and MySQL can use native change capture when configuration and permissions allow it, while other sources use scheduled snapshot refreshes.
This story is for Pro members
You've reached the end of the free preview. Upgrade to AlphaSignal Pro to read the full article - and everything else behind the paywall.
Comments
