Open Core
What's open, what's closed, and why.
Open Source
- Ingestion engine
- Data connectors (BLS, Census, FRED, WorldBank, 40+)
- Query layer and views system
- CLI tools
- API server (FastAPI)
- Schema inference
- Dataset YAML spec
- MCP server
Hosted Platform
- Web frontend
- Semantic search
- User accounts and profiles
- Starring, collections, follows
- Knowledge graph explorer
- LLM metadata enrichment
- Managed cloud infrastructure
Why this boundary
The split follows a simple principle: open source the infrastructure, keep the experience proprietary. Everything about how data gets in, how it gets queried, and how the system extends is open. Anyone can verify how data is handled, run the whole stack locally, or build on top of it.
The hosted platform adds discovery, collaboration, and managed infrastructure on top. That's where the business model lives. Researchers and journalists who need to trust the data pipeline can inspect every line of code. Organizations that want the full platform can use the hosted version.
What this means for you
Self-hosting: the open source core runs standalone with Docker Compose. Ingest data, query it, build views, all without the hosted platform. See the GitHub repo for setup instructions.
Contributing: connectors, aggregators, and views are all community-extensible. The entire extension model is declarative YAML, so adding a new data source doesn't require writing Python.