Multi-Tenant Architecture

Each customer organisation (tenant) gets a fully isolated environment - their data never mixes with another customer's data.

Tenant isolation is enforced at every level: subdomain routing, database separation, token validation, and background processing context.

Plans and AI usage

Each organisation is on a plan - Small Team, Pro or Enterprise - that sets how many users and shared workspaces it has and how much AI work it can do per month. AI work is measured in credits: chat conversations, discovery runs and document analysis use credits in proportion to the AI effort they take, so a quick question costs a few credits and a deep discovery run more. Small Team includes 5,000 credits a month, Pro 2,000 per user, Enterprise has no fixed limit. Administrators get a notice at 80 % of the budget. Once it is used up, new chats, discovery runs (including scheduled ones) and bulk analyses pause until the next billing period or an upgrade, and the user sees an upgrade prompt instead of an error. Everything already in the workspace stays readable.

Every organisation can also have one sample workspace with fictitious documents, findings and hypotheses to try things out. New trial teams get it automatically; existing organisations can create it on the Workspaces page. Every user automatically has access to it with write permissions, including colleagues who join later. It does not count towards the workspace limit, so it holds only a few documents (six in total, its demo files included) - real work belongs in a workspace of your own. Whoever created it can delete it at any time, after which a new one can be created.