schedulermcp is a modular Python library for scheduling tasks in AI agent pipelines. Cron jobs, intervals, one-shot delays, event-driven watch triggers. Webhook notifications, kernel execution, SQLite persistence.
from schedulermcp import SchedulerFactory
scheduler = SchedulerFactory.default()
scheduler.start()
await scheduler.schedule(
goal="Generate daily analytics report",
job_type="cron",
cron="0 9 * * 1-5",
webhook_url="https://api.example.com/reports",
)
await scheduler.schedule(
goal="Send welcome email in 5 minutes",
job_type="once",
delay_seconds=300,
)
from schedulermcp.mcp_server import SchedulerMCPServer
SchedulerMCPServer(scheduler).run()
5-field cron syntax — wildcards *, steps */N, ranges N-M and lists N,M. Recurring jobs with auto-computed next run.
Repeat tasks at fixed intervals with failure tracking and auto-pause after consecutive errors.
Fire-and-forget tasks with a delay. Run once after N seconds, or at an absolute datetime.
Execute scheduled jobs via kernelmcp. AI-powered task execution with full ReAct reasoning.
Send HTTP POST notifications when jobs fire. Track results, timing, and errors automatically.
Persist jobs across restarts. Zero-config file-based database. In-memory store also available.
9 MCP tools: schedule_task, list_schedules, get_job, cancel_schedule, pause_schedule, resume_schedule, and more.
SchedulerFactory.default() works without any external service. Add SQLite and webhooks as you scale.
One pip install. Zero dependencies to start.
Create jobs with goal, type, and timing. Scheduler computes next_run and persists.
Background loop checks for due jobs every tick. Executors handle the work.
Expose as MCP server, embed in your agent, or run standalone.
| Type | Description | Repeats |
|---|---|---|
| once | One-shot — fire once after delay or at absolute time | No |
| cron | Cron — recurring via cron expression | Yes |
| interval | Interval — repeat every N seconds | Yes |
| watch | Watch — event-driven, fires when shell command output matches condition | Yes |