Pricing
We read your library in at no charge, hold it in memory for a small monthly fee, and charge a flat rate per question. We never charge for input or context tokens.
Estimate your bill ↓Estimate your bill
Set your document count and your monthly questions to see the whole bill, meter by meter.
Two meters. Onboarding, input, and context are always free.
Input and context tokens are free at every volume. Onboarding is free too. You pay a small amount to hold each document in memory and a flat rate per question answered.
Plans
Start on your own, or reserve committed capacity for sensitive workloads.
Self serve, up and running the same day. You pay only for the two meters, with a small monthly minimum, and you can grow your library whenever you like. Best when you want to try it on a real workload without a commitment.
Committed capacity for regulated and data sensitive teams. Reserve steady performance for a large library and a busy question load, with the isolation and controls your compliance team expects. Best when memory holds material you cannot share.
Why it stays cheap
The work of reading a document happens a single time, at onboarding, and we do not charge for it. Growing your library costs nothing up front.
Because the document already lives in memory, answering a question never re-reads it. Each answer costs a fraction of a per token AI call, so a busy library stays affordable.
Holding a document in memory is inexpensive, so a large library stays affordable month to month. The bill grows gently with your library and stays predictable.
Performance
Response times stay steady as more people and agents ask questions at once. Because answers come from memory rather than re-reading, capacity keeps opening up while retrieval based AI slows down under the same load. Measured on a single dedicated serving tier.
Measured on a single dedicated serving tier as concurrent questions climb. Response times for Engram CaaS stay steady while retrieval based AI slows down, because every answer comes from memory instead of re-reading the source on each question.
