Second Brain

What is Second Brain?

Second Brain is a self-hosted memory layer for Claude and any MCP-compatible AI client. It gives your AI persistent, searchable memory across sessions using semantic search instead of keywords.

How to use Second Brain?

Deploy Second Brain with one click on Cloudflare’s free tier. Once deployed, it works with Claude Desktop, Claude Code, and claude.ai. No explicit installation commands or config keys are documented in the README.

Key features of Second Brain

• Five tools: remember, recall, list_recent, forget, append
• Semantic vector search by meaning, not keywords
• Duplicate detection blocks near-exact matches before storage
• Long notes are automatically chunked before embedding
• Built on Cloudflare Workers, D1, Vectorize, and Workers AI
• Runs entirely on the free tier with one-click deploy

Use cases of Second Brain

• Persistent memory across chat sessions for Claude
• Searchable knowledge base without exact keyword matching
• Tag-based chronological browsing of stored notes
• Updating previous entries with new timestamps
• Deleting outdated or incorrect memory entries

FAQ from Second Brain

How does Second Brain store and search memory?

Notes are embedded using bge-small-en-v1.5 on Workers AI and stored in Cloudflare Vectorize. Long notes are chunked before embedding so each section gets its own vector. Recall performs semantic vector search by meaning.

How does duplicate detection work?

Duplicate detection blocks near-exact matches before storing a new note. This prevents redundant entries when the same information is provided again.

What infrastructure does Second Brain require?

The stack uses Cloudflare Workers, D1, Vectorize, and Workers AI. Everything deploys automatically with one click and runs on Cloudflare’s free tier.

Which AI clients are supported?

Second Brain works with Claude Desktop, Claude Code, and claude.ai, as well as any MCP-compatible AI client.

What are the five tools and what do they do?

The tools are: remember (store a note with duplicate detection), recall (semantic vector search by meaning), list_recent (chronological listing by tag), forget (delete entry and all its chunks), and append (update an existing entry with a timestamp).