Skip to main content
Glama

⚠️ v2.0 breaking change: SQLite eliminado. v2 sólo soporta backend Obsidian. Datos en ~/.logbook/logbook.db (v1) no se migran automáticamente — quedan intactos en disco. Si necesitas acceso a ellos, mantente en @cocaxcode/logbook-mcp@0.4. Tools 10→5 (con shims deprecated). Ver CHANGELOG.md para detalles.


Quick Overview

logbook-mcp is an MCP server that turns your AI assistant into a persistent developer logbook. Capture decisions, track TODOs, set reminders, scan code TODOs, and search everything with full-text search — without leaving your editor.

It auto-detects your git project, stores everything locally, and works with any MCP-compatible client: Claude Code, Claude Desktop, Cursor, Windsurf, VS Code, Codex CLI, or Gemini CLI. All data stays on your machine — nothing is synced, nothing is tracked, nothing leaves your disk. Notes are scoped per-project automatically, but you can search globally across all your projects at any time.

Two storage modes: SQLite (default, zero config) or Obsidian (markdown files with frontmatter, visible in your Obsidian vault with Graph View, Dataview, Tasks, and Calendar). Switching from SQLite to Obsidian auto-migrates your data on startup.


Related MCP server: LumenCore

Just Talk to It

No commands to memorize. Just say what you need.

Capture notes

"I decided to use JWT instead of sessions for scalability"
→ Saved as a decision — retrievable months from now

"The CI is failing due to a timeout in integration tests"
→ Captured as a blocker — shows up when you review activity

Track TODOs

"TODO: implement email validation"
→ Created with auto-inferred topic

"Add these: fix token refresh. update deps. add rate limiting"
→ 3 TODOs created at once, each categorized

"Mark 5 and 8 as done"
→ Both completed

"What's pending across all projects?"
→ Global view of everything, including code TODOs

Set reminders

"Remind me tomorrow to deploy"
→ One-time reminder

"Remind me every Tuesday to review PRs"
→ Recurring weekly — auto-acknowledged after each session

"Remind me on weekdays to check the CI"
→ Monday to Friday

Search anything

"Why did we choose JWT?"
→ Finds the decision note, even months later

"Search everything about auth"
→ FTS5 search across all notes and TODOs

Installation

claude mcp add --scope user logbook -- npx -y @cocaxcode/logbook-mcp@latest --mcp

With Obsidian mode:

claude mcp add --scope user logbook -- npx -y @cocaxcode/logbook-mcp@latest --mcp --storage obsidian --dir "/path/to/vault/logbook"

Claude Desktop

Add to your claude_desktop_config.json:

{
  "mcpServers": {
    "logbook-mcp": {
      "command": "npx",
      "args": ["-y", "@cocaxcode/logbook-mcp@latest", "--mcp"]
    }
  }
}

With Obsidian mode:

{
  "mcpServers": {
    "logbook-mcp": {
      "command": "npx",
      "args": [
        "-y", "@cocaxcode/logbook-mcp@latest", "--mcp",
        "--storage", "obsidian",
        "--dir", "/path/to/vault/logbook"
      ]
    }
  }
}
  • macOS: ~/Library/Application Support/Claude/claude_desktop_config.json

  • Windows: %APPDATA%\Claude\claude_desktop_config.json

  • Linux: ~/.config/Claude/claude_desktop_config.json

Cursor — add to .cursor/mcp.json:

{
  "mcpServers": {
    "logbook-mcp": {
      "command": "npx",
      "args": ["-y", "@cocaxcode/logbook-mcp@latest", "--mcp"]
    }
  }
}

Windsurf — add to .windsurf/mcp.json:

{
  "mcpServers": {
    "logbook-mcp": {
      "command": "npx",
      "args": ["-y", "@cocaxcode/logbook-mcp@latest", "--mcp"]
    }
  }
}

VS Code — add to .vscode/mcp.json:

{
  "servers": {
    "logbook-mcp": {
      "command": "npx",
      "args": ["-y", "@cocaxcode/logbook-mcp@latest", "--mcp"]
    }
  }
}

Codex CLI:

codex mcp add logbook-mcp -- npx -y @cocaxcode/logbook-mcp@latest --mcp

Gemini CLI — add to .gemini/settings.json:

{
  "mcpServers": {
    "logbook-mcp": {
      "command": "npx",
      "args": ["-y", "@cocaxcode/logbook-mcp@latest", "--mcp"]
    }
  }
}

Configuration

logbook-mcp supports three ways to configure storage, with this priority order:

  1. CLI args (highest priority): --storage obsidian --dir "/path" --workspace "name"

  2. Environment variables: LOGBOOK_STORAGE, LOGBOOK_DIR, LOGBOOK_WORKSPACE

  3. Config file: ~/.logbook/config.json (auto-created on first run)

{
  "storage": "sqlite",
  "dir": null,
  "workspace": null,
  "autoMigrate": true
}

Switching from SQLite to Obsidian

Just change the config and restart. If autoMigrate is true (default), your existing SQLite data is automatically migrated to Obsidian on the next startup. No manual steps needed.

You can also check the current status with logbook_setup action:status.


Features

7 built-in topics

Every note, TODO, and reminder is categorized automatically by your AI, or you can specify a topic explicitly.

Topic

Purpose

Mapped from conventional commits

feature

New functionality

feat:

fix

Bug fixes

fix:

chore

Maintenance, CI/CD, refactoring

refactor: docs: ci: build: test:

idea

Future proposals

—

decision

Architecture choices

—

blocker

Things blocking progress

—

reminder

Time-based reminders

—

Custom topics can be created at any time — just say "create a topic called security".

Custom topics with type, folder, and dashboard

Topics can define their own behavior (kind), Obsidian folder, and whether they appear in the project dashboard (index.md):

  • kind: note (default) — each entry is an individual .md file

  • kind: todo — entries are checkboxes in a consolidated .md file

  • show_in_index: true (default) — adds a Dataview section and quick link to index.md

  • show_in_index: false — topic exists but is hidden from the dashboard

"Create a topic called incident with its own folder"
→ logbook_topics action:add name:"incident" kind:"note" folder:"incidents"
→ Entries saved to: project/incidents/2026-03-24-server-down.md
→ Dashboard updated with Incidents section

"Create a topic called sprint-task as todo type"
→ logbook_topics action:add name:"sprint-task" kind:"todo" folder:"sprint"
→ Entries saved to: project/sprint.md (as checkboxes)

"Create a private topic not shown in the dashboard"
→ logbook_topics action:add name:"internal" kind:"note" folder:"internal" show_in_index:false
→ Entries saved to: project/internal/ (not visible in index.md)

Without a folder, entries go to the default notes/ or todos/ directory. The dashboard (index.md) is auto-regenerated when a topic with show_in_index: true and folder is created.

Code TODO scanning

Your TODO, FIXME, HACK, and BUG comments are detected via git grep and shown alongside manual TODOs:

feature (2)
  [ ] #12 [manual] Implement email validation
      [code]  TODO: add OAuth support — src/auth/service.ts:45

fix (3)
  [ ] #8  [manual] Token doesn't refresh
      [code]  FIXME: handle null case — src/users/controller.ts:78
      [code]  BUG: race condition — src/chat/gateway.ts:112

When a code TODO disappears from source (because you fixed it), logbook detects it automatically and marks it as resolved.

Reminders

Both one-time and recurring patterns are supported:

Pattern

Example

Schedule

One-time

remind_at: "2026-03-25"

March 25 only

daily

Every day

Every day

weekdays

Monday to Friday

Mon–Fri

weekly:2

Every Tuesday

Specific day of week

weekly:1,3

Monday and Wednesday

Multiple days

monthly:1

1st of each month

Specific day of month

monthly:1,15

1st and 15th

Multiple days

Recurring reminders auto-acknowledge after being shown once per day. One-time reminders that were missed show as overdue.

Tip: logbook-mcp exposes an MCP Resource (logbook://reminders) that clients can load on session start. In Claude Code and Claude Desktop, reminders appear automatically without asking. In other clients, just say "any reminders?".

Full-text search (FTS5)

Search across all notes and TODOs instantly, powered by SQLite FTS5. Filter by topic, type, project, or search globally across all projects.

Batch operations

"Add: validate email. fix token. update deps"     → 3 TODOs created
"Mark 5, 8, and 12 as done"                       → 3 TODOs completed
"Delete TODOs 3 and 7"                             → 2 TODOs removed

Smart project detection

logbook-mcp auto-detects which git project you're in via git rev-parse. No config needed — it scopes queries to the current project by default, with a global option to see everything.


Tool Reference

Tool

Actions

Description

logbook_note

—

Add a note with optional topic

logbook_todo

add list done edit rm

Full TODO management

logbook_entry

list edit delete standup decision debug

Structured entries (ADRs, debug sessions, standups)

logbook_query

search log timeline

Full-text search, activity log, cross-project timeline

logbook_topics

list add

Manage topics (custom kind, folder, dashboard visibility)

logbook_tags

—

List tags with counts

logbook_reminders

—

View pending reminders

logbook_review

—

Weekly/monthly review with stats

logbook_inbox

list process

Quick notes inbox (Obsidian mode)

logbook_setup

init migrate status

Admin: init vault, migrate data, check status

10 tools + 1 resource

Param

Type

Required

Description

content

string

Yes

Note content (max 5000 chars)

topic

string

No

Topic name — AI infers it, or auto-created if new

action: add — Create TODOs

Param

Type

Description

content

string

Single TODO content (max 2000 chars)

items

array

Multiple TODOs: [{content, topic?, priority?, remind_at?, remind_pattern?}] (max 50)

topic

string

Topic — auto-inferred or auto-created

priority

low normal high urgent

Priority (default: normal)

remind_at

YYYY-MM-DD

One-time reminder date

remind_pattern

string

Recurring: daily, weekdays, weekly:N, monthly:N

action: list — List TODOs grouped by topic

Param

Type

Default

Description

status

pending done all

pending

Filter by status

topic

string

—

Filter by topic

priority

low normal high urgent

—

Filter by priority

source

all manual code

all

Manual DB or code comments

scope

project global

project

Current project or all

action: done — Mark as done / undo

Param

Type

Description

ids

number or number[]

ID(s) to mark

undo

boolean

If true, sets back to pending (default: false)

action: edit — Edit a TODO

Param

Type

Description

id

number

TODO ID to edit

content

string

New content

topic

string

New topic

priority

low normal high urgent

New priority

action: rm — Delete TODOs

Param

Type

Description

ids

number or number[]

ID(s) to delete permanently

action: standup — Daily standup

Param

Type

Required

Description

yesterday

string

Yes

What was done yesterday

today

string

Yes

What will be done today

blockers

string

No

Current blockers

topic

string

No

Topic

action: decision — Architecture Decision Record (ADR)

Param

Type

Required

Description

title

string

Yes

Decision title

context

string

Yes

Why this decision is needed

options

string[]

Yes

Options considered

decision

string

Yes

Decision taken

consequences

string

Yes

Consequences of the decision

topic

string

No

Topic (default: decision)

action: debug — Debug session

Param

Type

Required

Description

title

string

Yes

Bug/error title

error

string

Yes

Error description

cause

string

Yes

Root cause

fix

string

Yes

Solution applied

file

string

No

Attachment path

topic

string

No

Topic (default: fix)

action: list — List entries by type

Param

Type

Default

Description

type

note decision debug standup review

—

Entry type (required)

scope

project global

project

Current project or all

limit

number

20

Max results

action: edit / action: delete — Modify or remove entries by ID

action: search — Full-text search

Param

Type

Default

Description

query

string

—

Search text (required)

type

all notes todos

all

Search scope

topic

string

—

Filter by topic

scope

project global

project

Project or global

limit

number

20

Max results

action: log — Activity for a period

Param

Type

Default

Description

period

today yesterday week month

today

Quick date filter

from / to

YYYY-MM-DD

—

Custom date range

type

all notes todos

all

Filter by type

scope

project global

project

Current project or all

action: timeline — Cross-project timeline

Param

Type

Default

Description

period

today yesterday week month

week

Time range

workspace

string

—

Filter by workspace

action: status — Show current configuration and migration state

action: init — Initialize Obsidian vault (dashboard, templates, inbox)

Param

Type

Default

Description

force

boolean

false

Regenerate even if files exist

action: migrate — Manually migrate SQLite data to Obsidian (requires obsidian mode)


Storage

SQLite mode (default)

All data lives in a single SQLite database at ~/.logbook/logbook.db. Zero config.

  • WAL mode for concurrent reads

  • FTS5 virtual tables for instant full-text search

  • Triggers keep search indexes in sync automatically

  • Code TODO snapshots — tracks which code TODOs existed, detects when they disappear

Obsidian mode

Writes markdown files with YAML frontmatter directly to your Obsidian vault.

Configure via CLI args (recommended), config file, or env vars:

# CLI args (most reliable, especially on Windows)
claude mcp add logbook -- npx -y @cocaxcode/logbook-mcp@latest --mcp --storage obsidian --dir "/path/to/vault/logbook"
// Config file: ~/.logbook/config.json
{
  "storage": "obsidian",
  "dir": "/path/to/vault/logbook",
  "autoMigrate": true
}

Files are organized by workspace, project, and type:

vault/logbook/
├── cocaxcode/
│   ├── cocaxcode-api/
│   │   ├── notes/          ← logbook_note
│   │   ├── todos/          ← logbook_todo
│   │   ├── decisions/      ← logbook_entry action:decision
│   │   ├── debug/          ← logbook_entry action:debug
│   │   ├── standups/       ← logbook_entry action:standup
│   │   └── attachments/    ← copied files
│   └── cocaxcode-web/
├── optimus/
│   └── optimus-hub/

Each file has YAML frontmatter that Obsidian plugins can query:

---
type: todo
date: 2026-03-21
project: cocaxcode-api
workspace: cocaxcode
status: pending
priority: high
due: 2026-03-25
tags: [auth, urgent]
---
- [ ] Fix JWT refresh token

Recommended Obsidian plugins: Dataview (SQL-like queries), Calendar (date view), Tasks (checkbox management), Graph View (built-in, shows connections via [[wikilinks]]).

Auto-migration: When switching from SQLite to Obsidian, existing data is migrated automatically on startup (if autoMigrate: true). You can also run logbook_setup action:migrate manually.

Tip: Combine with Self-hosted LiveSync + CouchDB on your VPS to sync your vault across PC, Android, and iOS for free.


Architecture

src/
├── index.ts              # Entry: --mcp → server, else CLI
├── server.ts             # createServer() — 10 tools + 1 resource
├── config.ts             # Config resolution (args > env > file > defaults)
├── auto-migrate.ts       # Auto SQLite → Obsidian migration on startup
├── cli.ts                # CLI (help, version)
├── types.ts              # Shared interfaces
├── storage/
│   ├── types.ts          # StorageBackend interface
│   ├── index.ts          # getStorage() factory (uses resolveConfig)
│   ├── sqlite/           # SQLite backend (wraps db/)
│   └── obsidian/         # Obsidian backend (markdown + frontmatter)
├── db/                   # SQLite internals
├── git/                  # Git repo detection + code TODO scanning
├── resources/            # MCP Resource: logbook://reminders
└── tools/                # 10 MCP tools (one file each)

Stack: TypeScript · MCP SDK · better-sqlite3 · Zod · tsup


MIT · Built by cocaxcode

Related MCP Connectors

Related MCP Servers

  • A
    license
    Not graded
    quality
    D
    maintenance
    Provides AI coding assistants with persistent project memory to retain architectural decisions, code patterns, and domain knowledge across sessions. It stores data locally in a SQLite database, allowing agents to remember, recall, and manage project-specific context using full-text search.
    7 npm
    Apache 2.0
  • A
    license
    Not graded
    quality
    C
    maintenance
    Enables AI assistants to read and write to a personal knowledge vault of markdown notes, projects, and tasks, with tooling for search, capture, daily logs, and project management across different AI tools.
    MIT