HN Showcase
Back to browse
GitHub Repository

Brain-inspired persistent memory for Claude.ai. A single React artifact gives Claude salience scoring, forgetting curves, sleep consolidation, and context briefings — modeled on human hippocampal memory formation. Runs entirely inside the chat interface. No API key, no server, no extension needed.

32 starsTypeScript

Claude-engram – Brain-inspired persistent memory, runs inside Claude.ai

by mlapeter·Feb 16, 2026·6 points·2 comments
Visit ProjectView on HN

AI Analysis

●●●BangerBig BrainZero to One

Hippocampal memory model for Claude runs entirely in-browser, no API—genuinely novel architecture.

Strengths
  • •Insight is elegant: Claude artifacts can call Anthropic API + window.storage = persistent state machine without infrastructure
  • •Salience scoring, forgetting curves, sleep consolidation are actual neuroscience, not pretend—models real memory decay
  • •Setup is friction-free for power users: copy/paste React artifact, add one preference line, done
Weaknesses
  • •Requires manual memory dump→briefing cycle; not truly 'persistent' without user discipline between chats
  • •5MB storage ceiling and no guarantee window.storage persists; data loss risk if Claude session crashes
Category
AI/ML
Target Audience

Claude power users building long-term projects; researchers; anyone frustrated by Claude's session-reset amnesia

Similar To

LangChain memory modules · Mem0 · Anthropic memory features (built-in)

Post Description

Claude.ai artifacts can call the Anthropic API and have persistent storage (5MB via window.storage). I used these two capabilities to build a memory system modeled on how human memory actually works — salience scoring, forgetting curves, and sleep consolidation — all running inside a single React artifact with no external dependencies.

Just add artifact to your chat and paste instructions into your personal preferences setting.

Similar Projects

Developer Tools●Mid

CogmemAi – Persistent Memory for Claude Code via MCP

Runs extraction and search server-side so your local MCP is a tiny HTTP client — no local DBs, no giant RAM leaks, and an easy npx install and .mcp.json or global MCP registration. It exposes clear tools (save_memory, recall_memories, extract_memories, get_project_context) and adds project-scoped + global preferences — a pragmatic fix for Claude Code's tiny flat-file memory. The tradeoff is obvious: usefulness depends on the hosted API (privacy, uptime, cost), and the repo looks early-stage with minimal commits and docs beyond the quickstart.

Niche GemShip ItSolve My Problem
hifriendbot
204mo ago