Recordari Memory Tool Beats Microsoft Memora on LoCoMo Benchmark
corbymatt · reddit · 2026-08-14
The author benchmarked their memory tool, Recordari, against Microsoft's Memora using the open-source LoCoMo dataset harness. Recordari achieved an end-to-end accuracy of 0.831, outperforming Memora's published 0.801.
Key Details:
- The benchmark was run unmodified using Memora's original harness for fairness.
- Recordari scored 0.869 on multi-session questions, significantly beating the raw storage baseline of 0.760.
- The evaluation ran against the production API (MCP search and recall) rather than a mocked backend.
Compatible with Claude, ChatGPT, and other MCP-capable agents, the author has open-sourced the results and provided an anonymous sandbox for developers to test it themselves.
More from coding & agent
- Arcee Launches Open Models API Beta, Integrating DeepSeek and Kimi — latkins · 2026-08-14
- Arcee AI Open-Sources NAC, an Agent Harness for Long-Running Tasks — latkins · 2026-08-14
- Firecrawl Adds 41M+ Life Science Papers to Research Index for AI Agents — devdigest · 2026-08-14
- Vercel AI Gateway Natively Integrates Exa Search, Free Until August 31 — TheIshanGoswami · 2026-08-14
- Opinion: Overkill AI Skills Lead to Workslop Instead of Deep Work — Dupflo · 2026-08-14
- UnifyGTM CTO on Slashing 95% of Agent Costs Pre-Launch — hwchase17 · 2026-08-14