How do I build an AI second brain?

Capture daily notes, goals, and ideas in Basic Memory. Your agent helps structure them and links them into a personal knowledge graph it can read and write. Schedule the bundled memory-reflect skill (ships with the OpenClaw plugin) to consolidate recent notes periodically. Each morning, ask the agent what's on your plate. It walks the graph and assembles an agenda. The whole thing is plain Markdown you own and can edit in the web app or any editor.

Why not just use ChatGPT or Claude memory for a second brain?

Built-in AI memory is a small, vendor-held digest you can't fully read or export. It's good at short-term: the working thread, your preferences. A Basic Memory second brain is the long-term layer underneath: a large knowledge base of plain Markdown files you own, usable across every MCP-connected AI tool, not locked to one assistant. Both work together.

What does "periodic reflection" actually do?

The memory-reflect skill walks notes from the recent past, pulls out tagged observations, promotes one-off thoughts into project or goal notes where they belong, and writes a summary that links back to the underlying notes via `synthesizes` relations. Run it nightly, weekly, or by hand. The result: the brain consolidates instead of growing into noise.

Can I edit my second brain by hand?

Yes, and you should. Notes are plain Markdown files. Open it in the Basic Memory web app (or any editor), edit a daily note, correct what the agent wrote, add a relation by hand. Your edits sync back; the agent sees them on its next turn. You own the writes; the agent owns the writes too. Same files.

---

title: Build your second brain

type: playbook

for: [knowledge workers]

---

Build your second brain

The personal AI that actually remembers your life and work.

Everyone wants a Jarvis. Built-in memory will get you partway. We're the rest of the way. Plain-Markdown memory your AI reads and writes across goals, projects, decisions, and the things you don't want to lose. Three loops make it real: capture as you go, periodic reflection to distill, and a morning read where the agent surfaces what matters today.

Get Basic Memory Cloud

Start free

## The problem

AI alone can't hold all your context.

You want an assistant that knows your projects, your goals, and what you decided last month. Built-in AI memory forgets, and a notes app can't think. The middle, where the brain actually lives, is missing.

## How it works in Basic Memory

The workflow.

1Capture as you go. Open a daily note and brain-dump. Your agent helps structure it, tags observations like [goal], [decision], [idea], and links to existing projects and goals via wiki-links.
2Reflect periodically. Run the bundled memory-reflect skill (ships with the OpenClaw plugin) on a schedule. It consolidates recent notes into durable memory, promotes one-off thoughts into project or goal notes, and surfaces patterns across the week.
3Read the brain each morning. Ask the agent: 'what's on my plate today?' It runs search_notes, walks recent activity, and assembles an agenda from your tasks, calendar notes, and active goals, all from the same graph it has been writing to.
4Edit it from any device. Your daily notes, goals, projects, and decisions are plain Markdown files in your private cloud. Open the Basic Memory web app, read what your AI knows, and edit anything. The agent sees your changes on its next turn.

## What you get

The outcome.

A second brain that compounds. Daily notes link to goals; goals link to projects; projects link to decisions. The agent walks the graph each morning to find what's relevant today. Built-in memory still handles the working chat; Basic Memory holds the life graph underneath. Plain Markdown you can read, edit, and take anywhere.

## In practice

How a real second brain runs.

The pattern is three loops: capture as you go, reflect periodically, read each morning. Each loop touches the same plain-Markdown graph. Here's what the artifacts look like in practice. A daily capture note, the bundled reflection skill that consolidates them, and the morning agenda the agent assembles from the graph.

daily/2026-06-22.md (capture note, real)

Daily notes are the entry point. Brain-dumpy, fast, written during the day or at the end of one. Your agent helps structure them: questions become [question] observations, decisions become [decision], and wiki-links pick up real projects and goals. Over time the daily notes themselves become a graph of where the work went.

---
title: 2026-06-22
type: daily-note
tags: []
---

# 2026-06-22

- Who are our current top cloud users?
- What are they using BM for?
- What feedback do they leave when cancelling?

## Marketing

- Incoming traffic
  - Google — SEO keywords
  - GitHub — what links?
  - basicmemory.com / docs.basicmemory.com
- Conversion funnel on marketing site
- Checkouts that don't convert
  - WorkOS signup with no checkout
  - Failed/incomplete Polar checkout

## Content

- Walkthrough videos
- Testimonials
- How-to: codex, claude, cowork

## Observations

- [question] What's our actual cancellation reason mix?
- [todo] Add stats integrations for WorkOS, Polar, Umami, Google
  Adwords
- [idea] Onboarding flow asks "how did you find us?" — capture for
  channel attribution

## Relations

- relates_to [[Marketing Messaging & Site Gaps — June 2026]]
- progresses [[Goal 1.5 — Activation & Retention]]

memory-reflect (bundled skill, ships with OpenClaw plugin)

memory-reflect is one of the ten skills the OpenClaw plugin ships out of the box. It runs the periodic consolidation step: walk recent notes, promote one-off thoughts into the right project or goal note, and write a summary that links back via `synthesizes`. Schedule it with the scheduled-tasks MCP and you have a brain that consolidates on its own.

# What the agent does when memory-reflect runs

1. Find recent notes
   bm_recent(timeframe="7d")
   → 14 notes touched this week

2. Group by theme
   For each note: read tags, observations, relations.
   Cluster by project / goal / theme.

3. Promote where it belongs
   Loose [idea] observations from daily notes get appended to the
   project or goal they relate to. Daily notes stay as the source
   record; the promoted observation links back to the daily note
   it came from.

4. Write the weekly synthesis
   New note "weekly-summary-YYYY-WW" with:
   - what shipped (from project notes)
   - what's still open (from [todo] observations)
   - emerging patterns (from clustering)
   - "synthesizes" relations to every source note

5. Optionally schedule the next run
   Default: re-run weekly. Configurable to nightly or on demand.

morning (the agent assembling an agenda from the graph)

First thing in the morning, ask the agent. It runs a few BM tools, walks the graph, and produces a short agenda grounded in your actual goals, projects, and recent notes. No calendar integration required, just the graph the brain has already been writing. The agent reads from the same files you read; the agenda is just a fresh node on the graph.

# Morning agent flow

> What's on my plate today?

bm_recent(timeframe="3d")
  → daily notes, project updates, decisions touched recently

search_notes(tags=["todo"], status="in-progress")
  → 7 open todos across 3 projects

build_context(url="memory://goals/active")
  → current goals + the projects that progress them

# Agent answers:

Today's plate, drawn from the graph:

PROGRESS THE TOP GOALS
- [[Goal 1.5 — Activation & Retention]] — 3 todos open, last
  touched 2 days ago. Suggest finishing the WorkOS signup audit
  this morning.

DECISIONS THAT NEED A FOLLOW-UP
- 2026-06-22 daily note has a [question] about cancellation
  reasons. No follow-up note yet — flagging.

STALE
- [[Plan — Basic Memory on Rails]] hasn't been touched in 11 days.
  Either pick it back up or move it to /archive.

What it unlocks

A brain that compounds, in files you own

“Built-in memory keeps the chat. A notes app holds the words. Neither, alone, becomes a brain. The brain shows up when daily capture, periodic reflection, and a morning read all run on the same plain-Markdown graph. Each loop deposits something the next loop can build on. The brain compounds; you keep the file.”

We use this pattern ourselves. The daily note in the first artifact is real, from this workspace yesterday. The reflect and morning loops are running on the same graph.

## FAQ

Common questions.

How do I build an AI second brain?: Capture daily notes, goals, and ideas in Basic Memory. Your agent helps structure them and links them into a personal knowledge graph it can read and write. Schedule the bundled memory-reflect skill (ships with the OpenClaw plugin) to consolidate recent notes periodically. Each morning, ask the agent what's on your plate. It walks the graph and assembles an agenda. The whole thing is plain Markdown you own and can edit in the web app or any editor.
Why not just use ChatGPT or Claude memory for a second brain?: Built-in AI memory is a small, vendor-held digest you can't fully read or export. It's good at short-term: the working thread, your preferences. A Basic Memory second brain is the long-term layer underneath: a large knowledge base of plain Markdown files you own, usable across every MCP-connected AI tool, not locked to one assistant. Both work together.
What does "periodic reflection" actually do?: The memory-reflect skill walks notes from the recent past, pulls out tagged observations, promotes one-off thoughts into project or goal notes where they belong, and writes a summary that links back to the underlying notes via `synthesizes` relations. Run it nightly, weekly, or by hand. The result: the brain consolidates instead of growing into noise.
Can I edit my second brain by hand?: Yes, and you should. Notes are plain Markdown files. Open it in the Basic Memory web app (or any editor), edit a daily note, correct what the agent wrote, add a relation by hand. Your edits sync back; the agent sees them on its next turn. You own the writes; the agent owns the writes too. Same files.

Try this playbook.

Free and open source to run locally. Two minutes to connect your first AI tool.

Get Basic Memory Cloud

All playbooks

Build your second brain

AI alone can't hold all your context.

The workflow.

The outcome.

How a real second brain runs.

A brain that compounds, in files you own

Common questions.

Try this playbook.

Product

Resources

Community

Legal

#Build your second brain

#AI alone can't hold all your context.

#The workflow.

#The outcome.

#How a real second brain runs.

A brain that compounds, in files you own

#Common questions.

Try this playbook.

Product

Resources

Community

Legal

Build your second brain

AI alone can't hold all your context.

The workflow.

The outcome.

How a real second brain runs.

Common questions.