What is Helix Foundry?

Helix Foundry turns the tools your company runs on into one connected graph. Connect sources such as Stripe or your PostgreSQL database, send it your app’s or agent’s own data, upload the documents your tables don’t cover, let Foundry’s agent draft an ontology across them, then load it into a Helix database, ask it questions, and explore it as a graph or as records. Connectors keep syncing on a schedule, and ontologies refresh when their data changes. Foundry runs at foundry.helix-db.com, and your agents reach it through the Helix MCP. Sign in with the account you use for Helix Cloud; your workspaces come with it.

​

How it fits together

The work flows in one direction: You can do each step on the website, or have your agent do it through the Helix MCP, with your confirmation for every change.

​

What Foundry does not do

• It never writes to your sources. Connections only read, and database users need SELECT only.
• It never deletes a Helix database. Deleting an ontology removes what its loads wrote and leaves the database’s other data in place.
• It does not manage roles, member removal or single sign-on. Those stay in the Helix dashboard.

​

How Foundry uses AI models

A generation sends your synced tables, and up to five of the newest documents of each set it picks, to Foundry’s agent, which drafts the version. A load sends your documents’ text to AI models, which find the names and facts they hold and check them. Reading tables into a load uses no models. Ask sends your question, and what each step reads from the load, to AI models. By default these run on hosted providers through Helix, today Anthropic’s Claude models. On Pro or Max, an owner can run them on the workspace’s own Anthropic key instead; see AI models. What they cost counts against the workspace’s monthly caps.

​

Next steps