BEAM 100K
73.3%
Accuracy · 400 probes · 20 conversations
Both runs use the same test set.
BEAM test settings →Persistent memory for coding agents.
Keep project decisions, conventions, and fixes between sessions. Kimetsu stores them locally and retrieves the context your agent needs.
Benchmark results
All results and methodology →LoCoMo
89.4%
Accuracy · 1,540 questions
Questions about facts, events, and relationships across long conversations.
LoCoMo results and comparison →LongMemEval S
83.0%
Accuracy · 200-question stratified slice
Sampled from the 500-question set, covering recall, changed facts, and reasoning across sessions.
LongMemEval test settings →Historical results from the linked runs. An LLM answers and grades the questions; the memory pipeline uses local search, embeddings, and reranking, with no LLM calls.
How it works
Technical documentation →Store
Memories live in a SQLite database in your project. No hosted memory account is required.
Retrieve
Start with keyword search. Add local embeddings and reranking for semantic search. Set a token budget for the context sent to your agent.
Update
Track changed facts and suppress outdated results. Export and import memories to move them between machines.
Install Kimetsu
Select your agent and run these commands from your project directory. Available for Linux, macOS, and Windows.
All installation options →npm install -g kimetsu-ai
kimetsu setup --host claude-codeRequires npm. Installs the lean build; local semantic models are optional.
About Kimetsu
Created and maintained by Rodrigo Córdoba. Written in Rust and released under the MIT and Apache-2.0 licenses.
Contact
For questions, feedback, or collaboration, contact Rodrigo on LinkedIn.