FastRecall

What it is

Quick answers

Why FastRecall?

FastRecall solves the problem of being locked into a single AI model. With routers and multi-agent systems becoming essential infrastructure, context should move seamlessly between models and providers. Existing memory solutions are overcomplicated, slow, and expensive. FastRecall offers SOTA model-free compaction and lets the model organize its own context, making it faster, cheaper, and more performant.

Do you host models?

No, FastRecall does not call any model APIs on your behalf. You manage your own model integrations.

How is FastRecall so cheap?

FastRecall uses highly optimized and efficient context storage to deliver fast and free context recalls. You only pay for context storage on disk, the cheapest part of the stack. The platform does not use expensive LLM-based memory compaction techniques that introduce latency, cost, and retrieval errors.

What if I want to make use of model provider caching?

FastRecall is designed to work alongside existing model provider caching mechanisms. If you re-use the same model and context, you can leverage the existing cache for reduced token costs. See the Native continuation section of the documentation for more information.

Do I need to pass in the entire context into every model call?

No, FastRecall offers its SOTA model-free FlashCompact system to intelligently organize contexts over 1M tokens, or as an option flag in API calls. See the FlashCompact section of the documentation for more information.