gotoalberto

Second Brain Cc — AI skill for Claude Code

AI community

Persistent memory + working protocol for AI coding agents (Claude Code).

How to install Second Brain Cc

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

What Second Brain Cc does

Persistent memory + working protocol for AI coding agents (Claude Code). Numbered Markdown vault, stdlib SQLite FTS search, git sync, hooks/skills, optional 1Password + S3 modules. Clone and run one script.

Alternatives in AI

Peekaboo — Peekaboo is a macOS CLI & optional MCP server that enables AI agents to capture screenshots of applications, o 3.2k ★
Claude Skill Web Clone — Methodology for faithfully cloning any website (static / React / WebGL) — without copying AI-hallucinated code 969 ★
Session Kit — Close the terminal windows, not the AI coding sessions 649 ★

README

Second Brain

Persistent memory and a working protocol for **any AI agent** — an [Obsidian](https://obsidian.md)-compatible vault of Markdown notes plus a small, standard-library-only Python engine: a SQLite FTS5 search index and a git-synced repo. Notes survive across sessions and machines, and a lightweight protocol tells the agent how to search that memory, isolate its work, and write back what it learned.

It is **agent-agnostic**. The knowledge and the engine are plain files and stdlib Python; agents connect through one of three thin integrations — an MCP server, a CLI, or a native Claude Code layer. Swap the agent, keep the brain.

Why

Agents start every session with a blank slate. They forget the decision you made last week, the convention this repo follows, and who the people and systems in your world are. This project gives an agent two things:

Persistent memory — durable notes (decisions, conventions, how-tos, project state, people and systems) that outlive any single session.
A working protocol — a repeatable loop: search the vault for context first, work in isolation, then save what changed.

Three ways to connect your agent

The vault is the same for all of them; pick whichever your assistant speaks.

Integration	For	Setup
MCP server	Any MCP agent — Claude Desktop, Cline, Cursor, Continue, Zed, Windsurf, your own client	Point it at `integrations/mcp/server.py`
CLI	Any agent that can run a shell, and you at a terminal	Put `integrations/cli/brain` on your `PATH`
Claude Code	The deepest experience: automatic recall, agents, slash commands	`bash integrations/claude-code/install.sh`

All three drive the **same** `_bin/` engine, so redaction, locking, indexing and the frontmatter contract hold no matter which agent is writing. You can use more than one, but avoid running two write-back integrations at once.