strmforge/memcore-cloud

Memcore Cloud

<p align="center"> <img src="assets/brand/yifanchen-logo.jpg" alt="Memcore Cloud" width="220"/> </p> <p align="center"> <strong>Keep local AI agents from starting over.</strong> </p> <p align="center"> Memcore Cloud is a local AI memory layer for agents: capture source records, recall them with source refs, answer from evidence, install a standing agent rule, and check health before real recall. </p> <p align="center"> <a href="README.zh-CN.md">简体中文</a> · <a href="https://github.com/strmforge/memcore-cloud/releases/tag/v2026.6.20.1">2026.6.20.1</a> · <a href="LICENSE">MIT</a> </p> <p align="center"> <img alt="Version" src="https://img.shields.io/badge/version-2026.6.20.1-2f5f9b"> <img alt="Platforms" src="https://img.shields.io/badge/macOS%20%7C%20Linux%20%7C%20Windows-ready-247447"> <img alt="Local first" src="https://img.shields.io/badge/local--first-memory-b07d35"> </p>

Memcore Cloud is the English product name. 忆凡尘 / Yifanchen remains the Chinese name and codename.

Memcore Cloud is easiest to understand as a five-step local workflow:

capture -> recall -> answer from evidence -> install agent rule -> health

It is not a hosted chat app and not a summary-to-vector API. It keeps original records on your machine, returns source refs before raw excerpts, and gives local agents a standing rule for when to check memory before they answer or act.

Core Workflow

• Capture source records: keep original conversations, tool output, source tool, device, and timeline before any summary. Summaries help navigation, but original records remain the source of truth.
• Recall with source refs: ask about old decisions, preferences, fixes, project boundaries, or next steps and get compact source refs, library ids, hit reasons, and optional bounded excerpts.
• Answer from evidence: when a model is configured, Memcore Cloud can ask it to answer only from supplied evidence, cite supporting refs, or return `UNK