lifenote

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 login

That'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.

---

What this unlocks

Real things you can ask, once connected:

"I have a hard conversation with my cofounder today. What have I been writing about her?"

"I'm stuck. What was I trying to accomplish this month?"

"What did Marcus Aurelius (or Jung, or my Inner Wisdom) say to me about anxiety?"

"I remember saving a beautiful Jung line about words being vessels — find it."

"Pre-board-meeting check: what's been weighing on me this week?"

"What's my Life Goal, what's my focus this year, and what practices am I doing?"

"I'm drafting a post about late-bloomer entrepreneurship. What have I journaled on this theme?"

The agent figures out which tools to call. You just ask in plain English.

---

Install

# Recommended
pipx install lifenote

# Or with pip
pip install lifenote

Requires Python 3.10+.

Sign in

lifenote auth login

Browser 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.

For CI / Docker / headless agents, set LIFENOTE_API_TOKEN instead.

---

Use from your terminal

lifenote search "anxiety about launch"
lifenote entries --since 7d
lifenote today
lifenote week
lifenote conversations list
lifenote conversations show 574

Add --json to any command for piping into agents:

lifenote search "burnout" --json | claude "summarize the patterns"

---

Connect to your AI agent

Claude Desktop

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 Code

claude mcp add lifenote -- lifenote mcp

That's it. Open any Claude Code session and ask away.

Claude.ai web (Connectors)

Settings → Connectors → Add Connector → paste:

https://mcp.mylifenote.ai/mcp

Approve the OAuth screen. No install needed.

ChatGPT (Pro / Team / Enterprise / Edu)

OpenAI added remote-MCP-server support in late 2025. Steps:

1. Open ChatGPT → Settings → Apps & Connectors → Advanced settings
2. Toggle Developer Mode ON
3. Back in Apps & Connectors → Add new connector
4. Name: Life Note, URL: https://mcp.mylifenote.ai/mcp
5. Authentication: OAuth, then confirm "I trust this application"

Free-tier ChatGPT doesn't support remote MCP servers yet — paid plans only.

Gemini

Two paths, depending on which Gemini you use:

• Gemini CLI (Google's developer CLI): supports MCP natively. Configure lifenote as an MCP server in your Gemini CLI settings (~/.gemini/settings.json or similar) — same command: 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.

Hermes (NousResearch)

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: oauth

Then /reload-mcp inside any Hermes session. Full guide: docs/HERMES.md.

OpenClaw

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 --fix

Full guide (including OAuth bridge for desktop and bearer-token for headless): docs/OPENCLAW.md.

Cursor / Codex / other MCP clients

Same command: lifenote, args: [mcp] pattern, or point them at the HTTP endpoint directly.

---

Install the bundled Claude Code skill

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 install

Drops a markdown file into ~/.claude/skills/lifenote.md. Edit to taste.

---

What an agent can read

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.

---

Tools your agent can call

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 chats
• list_recent_entries(since?, limit?) — most recent entries with previews
• get_entry(entry_id) — full text + analysis + mentor responses

Wisdom — passages you saved (UI: "All Wisdom")

• list_wisdom(mentor?, since?, limit?) — your highlights, newest first
• search_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 entries
• list_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 Goal
• get_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.

---

CLI command reference

lifenote auth login | logout | whoami
lifenote search <query> [--since 30d] [--limit 20] [--json]
lifenote entries [--since 7d] [--limit 20]
lifenote today | week
lifenote conversations list | show <id>
lifenote mcp                  # stdio MCP server entry point
lifenote skills install       # bundled Claude Code skill
lifenote version

---

Configuration

Environment variables (all optional):

• LIFENOTE_API_TOKEN — bypass keychain (essential for CI / Docker)
• LIFENOTE_BASE_URL — point at staging / dev (default: https://www.mylifenote.ai)

---

How this is different from "just an API"

	"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.

---

Privacy

• 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.

---

License

MIT. Built by Daniel W. Chen. Source: https://github.com/dandanbang/lifenote-cli. Issues + PRs welcome.