shawnpetros

Open Brain

Name: Open Brain
Author: shawnpetros

Data community

SOURCE https://github.com/shawnpetros/open-brain

Description

Semantic memory for AI agents. Postgres + pgvector + MCP. Capture, classify, retrieve.

Installation

This entry records only its repository, not the path inside it, so there is no exact command to give. Open the source below and copy the folder into ~/.claude/skills/, or the file into ~/.claude/agents/.

README

Open Brain

**Because one brain is not enough in the age of the centaur.**

Your second brain, but make it semantic. A personal knowledge base that understands *meaning*, not just keywords — accessible from Claude, ChatGPT, Cursor, your phone, your terminal, or anywhere that speaks MCP.

You think a thought. Your brain captures it, figures out what kind of thought it is, who you mentioned, what topics it touches, and files it away with a vector embedding. Later, you ask a question in natural language, and your brain actually *understands* what you meant.

No folders. No tags. No "wait, where did I put that?" Just vibes. Semantic vibes.

What's Inside

you → "remind me about that conversation with Alex about the API redesign"
                              ↓
                    [semantic search] ← pgvector cosine similarity
                              ↓
brain → "Found 3 thoughts: Meeting with Alex on Feb 12..."

**7 tools, one brain:**

Tool	What It Does
`capture`	Save a thought — auto-generates embedding + extracts metadata
`semantic_search`	Find thoughts by meaning, not keywords
`search_by_person`	"What do I know about Sarah?"
`search_by_topic`	"What have I been thinking about system design?"
`list_recent`	"What's been on my mind this week?"
`stats`	"How active has my brain been?"
`delete_thought`	"Forget I said that."

Every thought gets automatically classified (decision, insight, idea, reflection...), people are extracted, topics are tagged, and action items are surfaced. All of this happens invisibly — you just think, it just files.

The Stack

No exotic dependencies. No "run these 47 Docker containers" energy. Just:

Neon Postgres + pgvector — your thoughts, vectorized and indexed (HNSW, cosine similarity, 1536 dimensions)
OpenAI — text-embedding-3-small for embeddings, gpt-4o-mini for metadata extraction
Next.js on Vercel — MCP ser