Legal
Terms of service
Last updated September 29, 2026
These terms govern use of Midium, an inference and agent-context service from RecursionAI LLC. By using the Services you agree to them.
Token billing
Midium token-billed inference is charged per million tokens at the rates published on the website. It does not retain prompts or generated responses. This tier is inference only and does not include Vaults. You are responsible for how you use model outputs.
Dedicated capacity
Dedicated capacity is hardware reserved for your team, with shared Vaults and no per-seat fee. Starter is $444/mo. Standard is $589/mo. Pro is $1,389/mo. Max is $2,149/mo. Ultra is coming soon. A Hub license is $200/mo for a Mac mini (64 GB memory or less) and $500/mo for a Mac Studio (more than 64 GB memory). Reserving a tier requires a deposit of one month's fee, invoiced after we confirm availability, and applied to your first invoice. Prompts and generated responses are not retained.
Hub license
A Hub license runs the same Vaults, sub-agents, and API on hardware you own and manage: $200/mo for a Mac mini (64 GB memory or less) and $500/mo for a Mac Studio (more than 64 GB memory), unless we agree otherwise in writing. It includes shared Vaults. Local inference stays on the licensed machine.
Midium Desktop
Midium Desktop is free. It includes one personal Vault. Shared Vaults come with dedicated capacity or a Hub license. Any model your hardware can run. Local inference and your personal Vault stay on your machine by default. Connectors only send what you point them at.
Vaults
A personal Vault is free in Midium Desktop. Shared Vaults come with dedicated capacity or a Hub license. There is no per-seat fee. Vault contents persist until you delete them. We do not train on vault contents. Cloud context is not used for training. Local Vaults stay on the device.
Acceptable use
You will not use the Services to violate law, attack systems you do not own, or attempt to extract weights or circumvent licensing.