Skip to main content
Guide
Foundry uses AI models to draft ontologies, read documents into loads, check the facts it finds, and answer questions. Settings → AI models says where they run. Owners choose; everyone else in the workspace sees the choice.
Settings, AI models: Choose where AI runs, with Default chosen and Claude, with your own API key

​
Default

Default runs on hosted providers through Helix, and you pay for what you use. Every workspace starts on it, and it is there on every plan. Today it runs on Anthropic’s Claude models.

​
Your own Anthropic key

On Foundry Pro or Max, an owner can run the workspace’s AI on the workspace’s own Anthropic API key instead, under the terms of its Anthropic account.
1

Choose Claude

Select Claude, paste an Anthropic API key, and choose a model for each task, or keep the recommended ones.
AI models with Claude chosen: the API key field and the model for each task
2

Test and save

Select Test and save and confirm. Foundry tests the key and models before it uses them: the page follows the test and says when it passed, or why it failed. Until it passes, the setting before stays in force.
From then on, new drafts, document reads and questions run on your key. Documents are read again on their next load, at your key’s cost. To go back, choose Default and select Use the default; it takes effect at once. Foundry keeps the key encrypted and never shows it again. Spend on your own key counts toward the workspace’s monthly caps all the same. On Free, choosing Claude says that bringing your own model key needs Foundry Pro, with a link to upgrade.

​
If the workspace leaves Pro

A workspace that moves to Free while it runs on its own key is paused: its drafts, document reads and questions are refused until it is on Pro or Max again, or an owner chooses Default. The page says so.

​
Who can change it

Only owners. Admins and members see where the workspace’s AI runs.

​
Next steps

Usage

This month’s AI spend, against its caps.

Plans and billing

Upgrade to Pro or Max.