What counts as a seat?
A human. Every person connected to the memory is a seat; every agent they run, personal or task-level, is free. We price this way deliberately: the value of a shared memory grows with the number of minds using it, and we never want you rationing agents to manage a bill.
Is there a free way to start?
Yes: Community is free forever, no card required, with unlimited agents, up to three connected tools and 90 days of memory history. Personal adds unlimited tools and full history, with the first month free. Teams typically start with a design-partner pilot: a real deployment on a real team, with success criteria we agree up front.
How does White Tree lower our AI costs?
Retrieval-based systems reprocess your archive on every question. White Tree recalls the relevant trajectory instead: the thread, not the corpus. Customers should expect falling per-task token costs as the memory fills, because every lesson captured is a reconstruction your agents never pay for again.
Who owns the memory?
You do, unambiguously. Your memory is your asset: it is never our training data, every record carries provenance and permissions you control, and you can export the entire memory at any time. If you leave, your judgement leaves with you.
Can we self-host or deploy privately?
Private deployment options are part of the Enterprise plan, and fully bespoke topologies, including federation with partner organisations, are what Heartwood exists for.
What happens when we switch models?
Nothing. Trajectories live below the model layer, so swapping Claude for next year's frontier model, or running five models side by side, carries the whole memory along untouched. Model churn is an event for your infrastructure team, not for your company's judgement.
