Build apps that answer, learn, and remember.
Engramic open source is a server framework for developing native memory based LLM Applications and APIs.
en·gram (ˈen-ˌgram, ˈēn-ˌgram)
Noun
A lasting trace of experience stored in the brain, theorized to be the physical basis of memory and learning.
-
Engramic is built to support proprietary input while continuously learning from interactions. It integrates and retains responses as structured, thoughtful memories—filtered, validated, and seamlessly incorporated into future outputs.
-
Engramic optimizes context management, ensuring retrieval and the LLM’s context window don’t just process data—they understand its relevance within the entire corpus. A structured pre-processing phase organizes data for maximum runtime efficiency, enabling smarter and more precise responses.
Repos and their associated memories* are instantly filtered by repo. Mix and match memories on the fly.
*Files are saved as memories.
Multi-Repo
Memory includes context and a timestamp. Can be used for episodic and procedural memories.
Persistent Memory
Stop filling your context window with more history than it needs. Engramic’s working memory is efficient and effective.
Working Memory
Licensing
Community (<250 Employees)
Open Source with the same terms as MIT plus a non-circumvent term.
Commercial (>=250 Employees)
Requires an affordable project-based commercial project license that scales with use.