Coming soon — join the waitlist

Your agents run
on your schedule

Durable cron jobs for AI agents. One API call to schedule. State persists between runs. Works with any framework via MCP.

✓ You're on the list.

POST /schedule

0 9 * * *

Cron fires

webhook called

Agent wakes up

state saved

Sleeps with context

sch_7kx9m2 — 200 OK · 1.2s

next run in 3h 41m

retry 2/3 — backoff 30s

state_size: 4.2kb

api.agentcron.live

// Schedule your agent in one call

POST /v1/schedules

{

"webhook": "https://my-agent.fly.dev/run",

"cron": "0 9 * * *",

"timezone": "America/New_York",

"state": { "last_checked": "auto" }

}

✓ Schedule created. Next run: tomorrow 9:00 AM ET

ID: sch_7kx9m2 · Retry: 3x · State: persistent

How it works

Three steps. That's it.

Point to your agent

Give us a webhook URL — wherever your agent runs. Fly.io, Railway, AWS Lambda, your own server. We don't care.

Set a schedule

Cron syntax, natural language, or interval. "Every weekday at 9am," "every 30 minutes," "once a week on Monday." Timezone-aware.

Your agent runs on time

We call your webhook on schedule with the state from the last run. Automatic retries on failure. Full execution log.

Features

Built for agents,
not humans

↺

Persistent state

Every run receives the state from the previous run. Your agent knows what it did last time without building your own storage layer.

⚙

MCP server

Your agent can schedule itself. Any MCP-compatible framework discovers AgentCron automatically. "Wake me up in 4 hours" just works.

↻

Smart retries

Exponential backoff with configurable retry count. If your agent's endpoint is temporarily down, we don't just give up.

⏱

Timezone-aware

Schedules run in your timezone, not UTC. No more converting "9am New York" to a cron expression in your head.

⚡

Instant webhook delivery

Sub-second invocation when your schedule fires. HMAC-signed payloads so your agent can verify it's really us.

☷

Execution dashboard

Every run logged with status, duration, response, and state diff. See exactly what happened and when it failed.

Use cases

What agents actually
need schedules for

0 9 * * *Morning inbox summary — scan email, Slack, and calendar. Send a digest before standup.

*/30 * * * *Lead monitoring — check CRM for new inbound leads every 30 minutes. Auto-qualify and route.

0 */4 * * *Competitor price tracking — scrape competitor pricing every 4 hours. Alert on changes.

0 22 * * *Overnight PR review — review open pull requests at 10pm. File issues by morning.

0 10 * * 1Weekly report generation — compile analytics data every Monday. Email the team a summary.

+3dFollow-up sequences — 3 days after first contact, agent sends a personalized follow-up.

MCP native

Your agent schedules itself

AgentCron ships with an MCP server. Any agent using Claude Code, CrewAI, LangChain, AutoGen, or any MCP-compatible framework can discover scheduling tools automatically.

Your agent doesn't need cron syntax or webhook configuration. It says what it needs in natural language, and the MCP tools handle the rest.

// Your agent's prompt:

"Check the support inbox

every 2 hours and escalate

anything marked urgent."

// Agent discovers AgentCron MCP

// and calls:

create_schedule({

interval: "2h",

task: "check_support",

persist_state: true

})

Pricing

Pay per invocation

Simple usage-based pricing

Free while in beta. Usage-based pricing at GA.

✓ Generous Free Tier

✓ Unlimited schedules

✓ State persistence included

✓ MCP server access

✓ 7-day execution log retention

✓ HMAC webhook signing