Blog & Skills

Notes, opinions, and copy-paste-ready skills.

stash: give your AI agent a 9-stage memory pipeline that learns across sessions

Every Claude Code session starts from zero. Stash is a self-hosted MCP server that consolidates your agent's raw observations into facts, relationships, contradictions, and patterns — memory that actually accumulates.

Sep 17, 2026 · 2 min read

TencentDB Agent Memory's L0-to-L3 semantic pyramid architecture diagram showing how raw conversations get distilled into atomic facts, scenario blocks, and user personas

skill · curated

TencentDB Agent Memory: the four-tier memory pyramid that stops your AI agent from starting fresh every session

Tencent's open-source memory system builds L0-to-L3 structured long-term memory on local SQLite — no external vector DB, no flat conversation dumps, no API bills.

Jul 27, 2026 · 2 min read