Bring your private Life Note journal into your AI agents — Claude Code, Claude Desktop, Cursor, Codex, anything that speaks MCP.
Life Note becomes the memory layer for your AI workflows. Every reflection you've ever written becomes context your agents can quietly reach for, so you don't have to re-introduce yourself.
pipx install lifenote
lifenote auth loginThat's it. From any Claude Code / Cursor / Claude Desktop session, you can now ask things like "what have I been writing about lately?" and the agent already knows you.
Once connected, just ask in plain English — the agent figures out which tools to call. Not sure where to start? Match it to what you need:
🌧️ Feeling down — "Look back through my journal and remind me of my small wins and the moments I actually grew — evidence I'm further along than I feel."
🔥 Angry or hurt — "I'm worked up about this. What have I written before about this person or situation — and what usually helps me settle?"
💡 Seeking clarity — "Across my recent entries, what are my top recurring patterns — and the triggers behind them?"
🌱 Looking to grow — "What blind spot keeps showing up in my writing — and what have my mentors said that speaks to it?"
🪧 Facing a decision — "I'm torn. Based on what I've written and my Life Goal, what do I actually value here?"
🧭 Feeling adrift — "What am I working on right now, and where am I drifting from this year's focus?"
The 14 read-only tools compose, so you can chain them:
"Cross-reference my journal and my saved wisdom on burnout — where do they agree, where do they diverge?"
"Pull my last month of entries and draft a weekly-review summary I can paste into my journal."
"Given my Life Goal and this year's focus, what have I actually made progress on — with evidence from my entries?"
# Recommended
pipx install lifenote
# Or with pip
pip install lifenoteRequires Python 3.10+.
lifenote auth loginBrowser opens, you click Allow, you're done. Your token is stored in your OS keychain (macOS Keychain / Windows Credential Manager / Linux libsecret) — never in plaintext.
A sign-in refreshes itself while you use it, but a weekly cron job needs something that just works. After lifenote auth login, create a personal access token:
lifenote token create --name "weekly obsidian sync" --expires 90d # 30d | 90d | 365d- It's read-only, and it is shown once. Life Note stores only a hash of it.
- Use it with
export LIFENOTE_TOKEN=ln_pat_…(cron, CI, Docker). With that set, the CLI never touches the keychain. Or add--saveto keep it in this machine's keychain as the default. lifenote token listshows each token's last use and expiry.lifenote token revoke <id>kills one at once.- A token can't create or revoke tokens. Requesting account deletion revokes them all.
Check what this machine will use:
lifenote auth status # who, which token, where it came from, days left| code | meaning |
|---|---|
| 0 | ok |
| 2 | usage: bad flag or value, or Life Note rejected the request as invalid |
| 3 | auth: not signed in, or the token is expired, revoked, or invalid |
| 4 | network: couldn't reach Life Note |
| 5 | server: Life Note returned an error, or rate-limited you |
Every failure prints one plain line to stderr.
lifenote search "anxiety about launch"
lifenote entries --since 7d
lifenote today
lifenote week
lifenote conversations list
lifenote conversations show 574Add --json to any command for piping into agents:
lifenote search "burnout" --json | claude "summarize the patterns"lifenote entries --since 7d --with-responses --json # analysis + every mentor reply per entry
lifenote entries --updated-since 2026-10-20T00:00:00Z --reply-after-id 98765 --all --json # only what changed
lifenote entries --after-id 12345 --all --json # only entries newer than id 12345--updated-since returns entries written or edited since that time, plus the other entries on those days, because a day's Daily Insight is shared and gets refreshed. --reply-after-id adds entries that got a new mentor reply. Every --json result carries as_of and reply_cursor: pass them back next time. lifenote sync does this for you.
lifenote export --since 2026-01-01 --until 2026-06-30 --format md --include-analysis --include-mentors -o journal.md
lifenote export --format csv --include-mentors -o journal.csv
lifenote export --since 30d --format json --include-analysislifenote sync <folder> keeps a folder of Markdown notes in step with your journal: one note per entry (2026/2026-10-28-12345.md), YAML frontmatter (date, mood, mentors, feelings, topics, people), the Daily Insight, and each mentor reply as an Obsidian callout.
lifenote sync ~/Obsidian/MyVault/"Life Note"- The first run pulls everything. Later runs fetch only entries written, edited, or given a new mentor reply since the last run, and refresh Daily Insights for those days. The cursor lives in the folder (
.lifenote-sync.json). - Re-running is safe: a note is rewritten only when it changed.
- Write your own thoughts below the
%% Your notes below this line… %%line in any note; sync keeps them. - Deleting an entry in Life Note does not delete its note.
--fullre-pulls everything.
Run it every week. Make a token once (lifenote auth login, then lifenote token create --name "weekly obsidian sync" --expires 365d), then:
cron (Linux / macOS): crontab -e, add:
# Sundays 08:00. Use the full path from `which lifenote`.
LIFENOTE_TOKEN=ln_pat_paste_yours_here
0 8 * * 0 /usr/local/bin/lifenote sync "$HOME/Obsidian/MyVault/Life Note" >> "$HOME/.lifenote-sync.log" 2>&1Exit code 3 in the log means the token expired or was revoked: make a new one.
launchd (macOS): save as ~/Library/LaunchAgents/ai.mylifenote.sync.plist, then launchctl load ~/Library/LaunchAgents/ai.mylifenote.sync.plist:
<?xml version="1.0" encoding="UTF-8"?>
<!DOCTYPE plist PUBLIC "-//Apple//DTD PLIST 1.0//EN" "http://www.apple.com/DTDs/PropertyList-1.0.dtd">
<plist version="1.0">
<dict>
<key>Label</key><string>ai.mylifenote.sync</string>
<key>ProgramArguments</key>
<array>
<string>/usr/local/bin/lifenote</string>
<string>sync</string>
<string>/Users/YOU/Obsidian/MyVault/Life Note</string>
</array>
<key>EnvironmentVariables</key>
<dict><key>LIFENOTE_TOKEN</key><string>ln_pat_paste_yours_here</string></dict>
<key>StartCalendarInterval</key>
<dict><key>Weekday</key><integer>0</integer><key>Hour</key><integer>8</integer><key>Minute</key><integer>0</integer></dict>
<key>StandardOutPath</key><string>/Users/YOU/.lifenote-sync.log</string>
<key>StandardErrorPath</key><string>/Users/YOU/.lifenote-sync.log</string>
</dict>
</plist>launchd runs a missed job when the Mac wakes; cron does not.
Edit Claude Desktop's config file and add the lifenote MCP server:
| OS | Config path |
|---|---|
| macOS | ~/Library/Application Support/Claude/claude_desktop_config.json |
| Windows | %APPDATA%\Claude\claude_desktop_config.json |
| Linux | ~/.config/Claude/claude_desktop_config.json |
{
"mcpServers": {
"lifenote": {
"command": "lifenote",
"args": ["mcp"]
}
}
}Restart Claude Desktop. Your journal is now a tool Claude can reach for on its own.
claude mcp add lifenote -- lifenote mcpThat's it. Open any Claude Code session and ask away.
Settings → Connectors → Add Connector → paste:
https://mcp.mylifenote.ai/mcp
Approve the OAuth screen. No install needed.
OpenAI added remote-MCP-server support in late 2025. Steps:
- Open ChatGPT → Settings → Apps & Connectors → Advanced settings
- Toggle Developer Mode ON
- Back in Apps & Connectors → Add new connector
- Name:
Life Note, URL:https://mcp.mylifenote.ai/mcp - Authentication: OAuth, then confirm "I trust this application"
Free-tier ChatGPT doesn't support remote MCP servers yet — paid plans only.
Two paths, depending on which Gemini you use:
- Gemini CLI (Google's developer CLI): supports MCP natively. Configure
lifenoteas an MCP server in your Gemini CLI settings (~/.gemini/settings.jsonor similar) — samecommand: lifenote, args: [mcp]shape as Claude Desktop. - Gemini Enterprise (Google Cloud paid product): supported via custom MCP server data stores.
- Gemini web app (
gemini.google.com) / Gemini mobile app: not yet supported by Google's regular Gemini app. Track this issue — Google is rolling MCP support out gradually.
Self-hosted autonomous agent across Telegram, Discord, Slack, WhatsApp, Signal, SMS, etc. Add to ~/.hermes/config.yaml:
mcp_servers:
lifenote:
url: "https://mcp.mylifenote.ai/mcp"
auth: oauthThen /reload-mcp inside any Hermes session. Full guide: docs/HERMES.md.
Self-hosted personal AI assistant ("the lobster way"). One-liner:
openclaw mcp set lifenote '{"url":"https://mcp.mylifenote.ai/mcp","transport":"streamable-http"}'
openclaw doctor --fixFull guide (including OAuth bridge for desktop and bearer-token for headless): docs/OPENCLAW.md.
Same command: lifenote, args: [mcp] pattern, or point them at the HTTP endpoint directly.
Teaches Claude Code when to reach for Life Note versus when not to (e.g. don't read your journal when you're debugging code):
lifenote skills installDrops a markdown file into ~/.claude/skills/lifenote.md. Edit to taste.
All read-only, scoped to your account by two permissions:
journal:read unlocks:
- Your written journal entries (with AI-generated analysis: feelings, topics, people, themes)
- Your Wisdom collection — passages you highlighted from mentor responses
- Your Practices (continuous goals) and Quests (deadline goals) — with streaks, subgoals, and the AI-distilled context (the practice, the wisdom anchoring it, the mentor message)
- Your Life Goal — your singular ultimate aim
- Your Yearly Goal — your chosen Theme + description for each year
- Journal entries you've explicitly linked to a specific practice or quest
chat:read unlocks:
- Your past conversations with AI mentors (Marcus Aurelius, Naval, your Inner Wisdom, etc.)
- The full message thread of any conversation you point to
What it can't do:
- Write to your journal, change your settings, see payment info, or access other users' data
- Cross-user data is structurally inaccessible — every query is scoped to your user ID
You can revoke any agent anytime at https://www.mylifenote.ai/settings/agents. Each connection has a separate revoke button.
Fourteen tools available via MCP, grouped by what they do:
Journal — your writing
search_entries(query, since?, limit?)— keyword search across entries, mentor responses, and chatslist_recent_entries(since?, limit?)— most recent entries with previewsget_entry(entry_id)— full text + analysis + mentor responses
Wisdom — passages you saved (UI: "All Wisdom")
list_wisdom(mentor?, since?, limit?)— your highlights, newest firstsearch_wisdom(query, limit?)— search within saved wisdom
Mentor wisdom — what mentors have said to you
search_mentor_wisdom(query, mentor?, since?, limit?)— across responses + chats, optionally filtered by mentor
Mentor conversations
list_conversations(since?, limit?)— past chats (mentor name + title + summary)get_conversation(conversation_id)— full message thread
Practices and Quests — what you're working on
list_practices_and_quests(status?, kind?, since?, min_streak?, has_active_subgoals?, limit?)— filter by active/completed, practice/quest, streak threshold, etc.get_practice_or_quest(item_id)— full goal + aligned_action context + subgoals + progress history + linked journal entrieslist_entries_for_goal(item_id, since?, limit?)— entries you tagged to this specific goal
Life direction — singular self-statements
get_life_goal()— your singular ultimate Life Goalget_yearly_goal(year?)— chosen Theme + description (default: current year)list_yearly_goals()— all years' themes (see how focus has evolved)
When you connect through Claude Desktop or Claude Code, all of these become available automatically. When you connect through Claude.ai's web Connectors, you'll see them in the conversation's tool list.
lifenote auth login | logout | whoami | status
lifenote token create --name NAME [--expires 30d|90d|365d] [--save]
lifenote token list | revoke <id>
lifenote search <query> [--since 30d] [--limit 20] [--json]
lifenote entries [--since 7d] [--until DATE] [--limit 20] [--with-responses]
[--updated-since ISO] [--after-id N] [--all] [--json]
lifenote export [--since] [--until] [--format md|json|csv] [--include-analysis]
[--include-mentors] [-o FILE]
lifenote sync <folder> [--full]
lifenote today | week
lifenote conversations list | show <id>
lifenote mcp # stdio MCP server entry point
lifenote skills install # bundled Claude Code skill
lifenote version
Environment variables (all optional):
LIFENOTE_TOKEN— a personal access token; bypasses the keychain (cron, CI, Docker).LIFENOTE_API_TOKENstill works.LIFENOTE_BASE_URL— point at staging / dev (default:https://www.mylifenote.ai)
| "Search your journal" | Life Note as memory layer | |
|---|---|---|
| What you ask | Specific queries | Conversational, contextual |
| What the agent knows about you | Nothing — you re-introduce yourself each session | Already knows your themes, your Life Goal, your saved wisdom |
| Result | Lookups | Calibrated responses |
The moat isn't the API. It's the years of writing already in your journal.
- All journal data is encrypted at rest with a per-user key.
- The CLI never stores plaintext tokens on disk — only in your OS keychain.
- Every agent connection is auditable at /settings/agents.
- We never train models on your data. We never share it with third parties.
- You can disconnect any agent (or all of them) with one click.
MIT. Built by Daniel W. Chen. Source: https://github.com/dandanbang/lifenote-cli. Issues + PRs welcome.