A local memory tool for Hermes agents
Engram Lite is a for . It lets an agent save and search memories on the user's own machine. Memory lookup does not call an LLM, so it aims to avoid hidden prompts and extra tokens just to retrieve past information.
In a judged by , it is presented as performing better than many existing s that mostly depend on LLM calls. The code is available on GitHub at engrammemory-labs/engram-lite.
Key points
- Engram Lite adds a for .
- It stores and searches memory on the user's own machine.
- It avoids using an LLM for memory lookup, which may reduce token use.
- It was tested with a using as judge.
- The GitHub project is engrammemory-labs/engram-lite.