claude-cron

claude-cron

MCP server that allows AI coding assistants to schedule deferred tasks such as reminders, shell commands, and AI prompts, executing them as an OS daemon.

Category
访问服务器

README

Claude Cron

License: MIT Bun TypeScript

Deferred tasks for AI coding assistants. MCP server + daemon that allows any AI coding assistant to schedule tasks for the future.

How does this compare to /loop? Claude Code 2.1.71 added /loop for in-session recurring tasks. It's great for ephemeral monitoring but dies when you close the terminal. claude-cron runs as an OS daemon — your tasks survive reboots and execute while you sleep. Read the full comparison.

Quickstart

bun install
bun test          # 38 tests
bun run typecheck # tsc --noEmit

Claude Code Configuration

Add to ~/.claude/settings.json:

{
  "mcpServers": {
    "claude-cron": {
      "command": "bun",
      "args": ["run", "/Users/fernando/code/claude-cron/src/index.ts"],
      "env": {
        "CLAUDE_CRON_DB": "/Users/fernando/.claude/claude-cron.sqlite",
        "GMAIL_ADDRESS": "frr@keepcoding.io",
        "GMAIL_APP_PASSWORD": "op://FRR DEV/...",
        "NTFY_TOPIC": "claude-cron-frr-xxx",
        "NTFY_URL": "https://ntfy.sh"
      }
    }
  }
}

Architecture

AI Assistant (Claude Code / Codex / Copilot)
    │ stdio (MCP protocol)
    ▼
MCP Server (src/index.ts)          ← Task CRUD
    │ SQLite (WAL mode)
    ▼
Runner / Daemon (src/runner.ts)    ← Loop: executes due tasks
    ├─ reminder  → notification + email + push
    ├─ shell     → Bun.spawn(command)
    ├─ claude    → claude -p --output-format text "prompt"
    ├─ codex     → codex --quiet "prompt"
    └─ copilot   → gh copilot suggest "prompt"

MCP Tools

Tool Description
schedule_task Schedule a task for the future
list_tasks List tasks (filter by status, upcoming, tags)
cancel_task Cancel a task by ID
run_now Mark a task for immediate execution
install_daemon Register the runner as an OS service

schedule_task

{
  name: "Nightly tests",
  type: "shell",              // reminder | shell | claude | codex | copilot
  when: "daily at 3:00",      // see "When expressions" below
  payload: {
    command: "cd ~/code/bfclaude && make test",  // shell
    // message: "...",         // reminder
    // prompt: "...",          // claude/codex/copilot
    // model: "haiku",
    // workingDir: "/path",
    // maxBudgetUsd: 0.10,
  },
  notify: ["notification", "email", "push"],
  notifyTo: "frr@keepcoding.io",
  tags: ["ci"],
}

when Expressions

Type Examples
Natural language "tomorrow 9:00", "in 30 minutes", "next friday at 3pm"
Cron (5 fields) "0 9 * * 1", "*/5 * * * *"
ISO 8601 "2026-03-01T09:00:00Z", "2026-03-01"
Shortcuts "every monday at 8", "daily at 9:00", "every 5 minutes"

Parsing: chrono-node (natural) + croner (cron) + regex (shortcuts). Shortcuts are evaluated before chrono-node so that "every monday at 8" is not interpreted as a single date.

Structure

src/
├── index.ts              # MCP server entry point (5 tools)
├── runner.ts             # Daemon: oneshot (launchd) or --loop (dev)
├── queue.ts              # SQLite: schema, CRUD, getDueTasks
├── scheduler.ts          # parseWhen(): chrono-node + croner + shortcuts
├── adapters/
│   ├── types.ts          # AIAdapter interface
│   ├── claude.ts         # claude -p --output-format text
│   ├── codex.ts          # codex --quiet (stub)
│   └── copilot.ts        # gh copilot suggest (stub)
├── notify/
│   ├── types.ts          # NotificationAdapter interface
│   ├── desktop.ts        # node-notifier (cross-platform)
│   ├── email.ts          # Direct SMTP to Gmail (TLS 465)
│   └── push.ts           # ntfy.sh (mobile + Apple Watch)
└── daemon/
    ├── types.ts          # DaemonInstaller interface
    ├── detect.ts         # process.platform → installer
    ├── launchd.ts        # macOS: ~/Library/LaunchAgents/
    ├── systemd.ts        # Linux: ~/.config/systemd/user/
    └── schtasks.ts       # Windows: Task Scheduler

Runner

Two modes:

# Oneshot (for launchd StartInterval / systemd timer)
bun run src/runner.ts

# Continuous loop (for dev or KeepAlive)
bun run src/runner.ts --loop

The daemon reads SQLite every 60s (configurable via CLAUDE_CRON_INTERVAL), executes due tasks, and sends notifications.

Cron tasks are automatically rescheduled after completion (next_run recalculated). One-off tasks transition to completed or failed.

Daemon

# Install (from Claude Code via MCP tool install_daemon, or manually):
# macOS:
bun run src/daemon/launchd.ts  # generates plist + launchctl load

# Verify
launchctl list | grep claude-cron
tail -f ~/Library/Logs/claude-cron.log

# Uninstall
launchctl unload ~/Library/LaunchAgents/com.claude-cron.runner.plist

Notifications

Channel Requirements Env vars
Desktop node-notifier (included)
Email Gmail App Password GMAIL_ADDRESS, GMAIL_APP_PASSWORD
Push (ntfy) ntfy app on mobile/watch NTFY_TOPIC, NTFY_URL (opt), NTFY_TOKEN (opt)

SQLite Schema

tasks (
  id TEXT PRIMARY KEY,          -- nanoid 12 chars
  name, type, status,           -- type: reminder|shell|claude|codex|copilot
  run_at INTEGER,               -- Unix ts (one-off)
  cron TEXT,                    -- Cron expression (recurring)
  payload TEXT,                 -- JSON per type
  notify_via TEXT,              -- JSON array: ["notification","email","push"]
  notify_to TEXT,               -- Destination email
  next_run, last_run,           -- Timestamps
  run_count, last_result,       -- Executions + JSON result
  created_at, created_by, tags  -- Metadata
)

Dependencies

  • @modelcontextprotocol/sdk — MCP protocol
  • chrono-node — Natural language → Date
  • croner — Cron parsing + next run
  • nanoid — Short IDs
  • node-notifier — Cross-platform desktop notifications
  • zod — MCP input validation
  • bun:sqlite — built-in, zero deps

TODO

  • [ ] Configure MCP in real settings.json and test end-to-end
  • [ ] Install daemon with install_daemon and verify full cycle
  • [ ] Set up ntfy for push to mobile/watch
  • [ ] npm publish (name TBD)
  • [ ] CI with tests on macOS + Linux
  • [ ] README with real usage GIFs

推荐服务器

Baidu Map

Baidu Map

百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。

官方
精选
JavaScript
Playwright MCP Server

Playwright MCP Server

一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。

官方
精选
TypeScript
Magic Component Platform (MCP)

Magic Component Platform (MCP)

一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。

官方
精选
本地
TypeScript
Audiense Insights MCP Server

Audiense Insights MCP Server

通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。

官方
精选
本地
TypeScript
VeyraX

VeyraX

一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。

官方
精选
本地
graphlit-mcp-server

graphlit-mcp-server

模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。

官方
精选
TypeScript
Kagi MCP Server

Kagi MCP Server

一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。

官方
精选
Python
e2b-mcp-server

e2b-mcp-server

使用 MCP 通过 e2b 运行代码。

官方
精选
Neon MCP Server

Neon MCP Server

用于与 Neon 管理 API 和数据库交互的 MCP 服务器

官方
精选
Exa MCP Server

Exa MCP Server

模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。

官方
精选