Open-source agent memory tool claims 94.7% on LongMemEval
intellinker · reddit · 2026-07-26
An open-source agent memory tool claims to keep context for weeks by adding one algorithm, rather than relying on opaque cloud-based memory systems.
- The author says the tool is designed for long-running agents and multi-topic research workflows.
- It uses a Laiden algorithm plus semantic graph-based engines to build node clusters that agents can retrieve later.
- The project is open-source and MIT-licensed, with PRs welcomed.
- The repo claims to beat mem0 and supermemory on LongMemEval, scoring 94.7%.
Repo: kunal12203/swafra
More from coding & agent
- Agent Skills are loaded in one prompt, with caching reducing multi-turn token cost — dotey · 2026-07-26
- Grok Build Web UI adds a local-first dashboard for live CLI sessions — elonmusk · 2026-07-26
- LLM config tuning thread weighs KV cache tradeoffs in an 80K-context Laguna setup — kingo86 · 2026-07-26
- OpenAI rolls out ChatGPT Voice on desktop with voice control for multiple agents — steipete · 2026-07-26
- Agensis adds a collaborative workspace for human-agent projects on your own hardware — jasonkneen · 2026-07-26
- Codex backs up old high-school hard drives and resurfaces deeply awkward files — shakoistsLog · 2026-07-26