context-vault and Structured-Memory-Engine
About context-vault
fellanH/context-vault
Persistent memory for AI agents — save and search knowledge across sessions via MCP. Local-first, markdown + SQLite + embeddings.
Implements hybrid full-text and semantic search via embeddings, with MCP tools for saving structured entry types (insights, decisions, patterns) and ingesting external content from URLs or projects. Runs as an auto-configured shared daemon that detects Claude, Cursor, and other AI tools, storing all data as plain markdown in `~/vault/` with SQLite indexing for search and optional web dashboard access.
About Structured-Memory-Engine
Bryptobricks/Structured-Memory-Engine
Persistent, self-maintaining memory for AI agents. 990 tests. <1ms recall. $0/month forever.
Implements a 6-signal ranking pipeline (keyword match + semantic similarity + recency + type priority + file weight + entity overlap) over SQLite FTS5 with local embeddings, enabling sub-50ms context injection without API calls. Features entity graph linking, confidence decay with configurable half-life, contradiction detection with temporal awareness, and query intent classification that surfaces action items or factual results based on question type. Ingests meeting transcripts into tagged markdown, auto-captures decisions from conversation, and includes a built-in recall benchmark suite to regression-test retrieval quality.
Related comparisons
Scores updated daily from GitHub, PyPI, and npm data. How scores work