- Home
- Skills
- Agents & Orchestration
- Second Brain Memory
More screenshots
Works with the AI Agents you already use
Second Brain Memory
Give your coding agent persistent memory it actually uses: recall past decisions before acting, save only what is worth keeping, and hand work between agents. Plain markdown notes you own, on a open-source MCP server.
Free
Second Brain Memory
Example session with this skill installed
Remember that we chose SQLite over Redis for the cache because it needs no ops. Use agent name demo.
What did we decide about the cache, and why?
- Read your context and instructions
- Compiled the second brain memory
Session 1: saving the decision
The agent called brain_remember and returned the note id:
agents/demo/Decision - Use SQLite over Redis for the cache
Session 2: a fresh session, no shared context
The agent called brain_recall and answered:
We decided to use
SQLite over Redis for the cache because it requires no separate server to run, configure, or maintain.
What is stored on disk
A plain markdown file in your vault folder (open it in Obsidian):
---
type: decision
tags: [database, cache, sqlite, redis]
agent: demo
createdAt: 2026-11-03
importance: 5
version: 1
---
# Decision - Use SQLite over Redis for the cache
We chose SQLite over Redis for the cache because it needs no ops: no separate server to run, monitor or maintain.
Tools involved
| Tool | Used for |
|---|---|
brain_remember | Save a new note |
brain_recall | Find relevant notes and follow their links |
brain_node_update | Append new facts to an existing note instead of duplicating it |
Next steps
- Ask your agent to recall context at the start of your next task
- Open the vault folder in Obsidian to see everything it remembers
- Read
workflows/02-remember-well.mdfor what is worth saving
Connects securely to your tools. The creator never sees your data.
About this skill
Agents forget everything between sessions. This skill gives yours a disciplined memory habit on top of a free, open-source MCP memory server, so it stops re-deriving decisions and asking you the same questions.
What your agent learns to do
- Recall before acting. Look up past decisions, constraints and fixes first, instead of starting from zero.
- Remember what is reusable. Save a decision with its reason, a hard constraint, a non-obvious fix, a convention. Not status updates or raw output. Search first and extend an existing note rather than creating duplicates.
- Hand off cleanly. Use a shared task queue (with dependencies) and agent-to-agent messages, so work moves between agents without long transcripts.
Your memory is a folder of markdown files
Open it in Obsidian, read exactly what your agents "know", edit or delete anything, back it up like any other folder. Notes link to each other with [[wikilinks]], and recall follows those links.
What is included
SKILL.md: the memory loop and the rules your agent follows- 6 step-by-step playbooks: starting a session, deciding what to save, decision records, incident notes, onboarding to a project, agent handoff
- A starter vault of example notes (projects, decisions, incidents, conventions)
SETUP.mdand an MCP config template
Setup (about two minutes)
- Add the free memory server:
claude mcp add second-brain -- npx -y mind-mri-second-brain(other clients: paste the config frommcp-config.template.json; on native Windows seeSETUP.md) - Add this skill to your agent
- Ask the agent to run
brain_health
Requirements
Node.js 22.5 or newer and an MCP-capable agent. Tested with Claude Code on Windows.
Transparency
- This package contains instructions and example notes only. The memory server is a separate, free, open-source (MIT) package by the same author: https://github.com/ara-5/mind-mri-agent-second-brain
- The server listens on localhost only and makes no outbound network calls unless you add optional API keys.
- The skill runs no shell commands and sends nothing anywhere.
Find more skills like this on the Claude Skills Marketplace, the Codex Skills Marketplace and the Gemini Skills Marketplace, or browse Claude Skills for Agents & Orchestration.
Needs Node.js 22.5 or newer; older versions will not run the memory server.
Tested only with Claude Code on Windows. Other MCP-capable agents and systems should work but have not been tested by the author.
The memory server is a separate free package. The setup command downloads it from npm the first time, so you need an internet connection for that step.
Notes live in a folder on one machine. Syncing between machines is up to you (for example by syncing the folder yourself); built-in git sync exists but is off by default.
Semantic (meaning-based) search needs your own OpenAI or Gemini API key. Without one, recall still works using the note links and keyword search.
On native Windows PowerShell 5.1, the claude mcp add command can drop the "--" separator. Run it from cmd.exe instead (SETUP.md explains this).
The skill guides the agent's habits, but cannot force them. Agents can still skip saving or recalling, so you may need to remind yours at first.
Do not save passwords or API keys as notes. The notes are plain text files and nothing is encrypted or scrubbed automatically.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before buying a skill.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
Fresh listing
Recently published to Agensi
Free to use
No account required to browse
Trust & safety
Security scan
Security scan passed 11 days ago
- Free with an Agensi account
- Software
- second-brain MCP
Node.js 22.5+mind-mri-second-brainConfig MCP
Git
The MCP server is free and open source (npm package mind-mri-second-brain).
Permissions
Allowed Hosts
File Scopes
The skill instructs the agent to use the Second Brain MCP tools (recall, search, remember, update, tasks, messages). The separate MCP server stores notes as markdown files in one folder (default ~/.second-brain/vault, or a folder you choose with BRAIN_VAULT) and listens on localhost only. No accounts or API keys are required. Optional, off by default: semantic search with your own OpenAI or Gemini key, and git sync of the notes folder. The setup command (npx) downloads the server once from the npm registry.