Back to browse
GitHub Repository

Bi-temporal agent long-term memory: SQLite-backed MCP server with recall ranking, hybrid vector+FTS recall, and a source-staleness signal

1 starsTypeScript

Memharness – Bi-temporal memory for AI agents, in one SQLite file

by sakuraiben·Jun 18, 2026·1 point·1 comment

AI Analysis

●●●BangerBig BrainZero to One

Bi-temporal facts with provenance beat vector-store bags for agent audit trails.

Strengths
  • Bi-temporal storage tracks when facts became true vs when agent learned them separately.
  • Zero LLM calls in storage layer means deterministic, auditable SQLite file.
  • Supersession not deletion enables 'what did you believe before I corrected you' queries.
Weaknesses
  • MCP server integration limits adoption to agents using that protocol specifically.
  • No built-in embedding generation means you handle vectorization separately.
Category
Target Audience

AI agent developers building long-term memory systems

Similar To

mem0 · Zep · Letta

Similar Projects

AI/ML●●Solid

Memv – Memory for AI Agents

Predict-calibrate extraction reduces noise, but Zep and Mem0 already dominate the agent memory space.

Big BrainNiche Gem
brgsk
432mo ago
AI/ML●●Solid

Prism MCP – The Mind Palace for AI Agents (Local-First Memory)

SQLite F32_BLOB vector search with Git drift detection for agent memory.

Eye CandyWizardry
dcostenco
102mo ago