⏰ LangGraph Lite Cron

• A free, self-hosted alternative to LangGraph Platform's Cron Jobs with the same API interface (compatible with langgraph-sdk).
• Designed for use with ambient agents to enable scheduled and recurring tasks.

Requirements: Python 3.9+, PostgreSQL, Redis

⚠️ Warning: Early development. Not for production use.

⚡ Quick Start

1. Install

pip install langgraph-lite-cron

Or, if you are using uv (including pre-release versions):

uv add --prerelease allow langgraph-lite-cron

2. Add Custom Routes

Add cron routes to your FastAPI app (e.g., ./src/agent/webapp.py):

from collections.abc import AsyncGenerator
from contextlib import asynccontextmanager

from fastapi import FastAPI

from langgraph_lite_cron import crons
from langgraph_lite_cron.scheduler import create_scheduler


@asynccontextmanager
async def lifespan(app: FastAPI) -> AsyncGenerator[None]:
    async with create_scheduler() as scheduler:
        await scheduler.start_in_background()
        app.state.scheduler = scheduler

        yield

        await scheduler.stop()
        await scheduler.wait_until_stopped()


app = FastAPI(lifespan=lifespan)
app.include_router(crons.router)

3. Configure langgraph.json

Update your langgraph.json to include the custom app:

{
    "dependencies": [
        "."
    ],
    "graphs": {
        "agent": "./src/react_agent/graph.py:graph"
    },
    "env": ".env",
    "image_distro": "wolfi",
    "http": {
        "app": "./src/agent/webapp.py:app"
    },
    "dockerfile_lines": [
        "ENV UV_PRERELEASE=allow"
    ]
}

4. Deploy

Option 1: In-memory mode (development)

langgraph dev

Option 2: Postgres + Redis mode (development)

langgraph up

Option 3: Self-hosted Standalone Server

• Deploy using Self-hosted Standalone Server with Kubernetes, Docker and Docker Compose.

5. Use Cron API

from langgraph_sdk import get_client

client = get_client(url="http://localhost:8123")

# Create thread
thread = await client.threads.create()

# Schedule job for thread (stateful)
cron_job = await client.crons.create_for_thread(
    thread["thread_id"],
    "agent",
    schedule="27 15 * * *",
    input={"messages": [{"role": "user", "content": "What time is it?"}]},
)

# Schedule job without thread (stateless)
cron_job_stateless = await client.crons.create(
    "agent",
    schedule="0 9 * * *",
    input={"messages": [{"role": "user", "content": "Good morning!"}]},
)

# List all cron jobs
crons = await client.crons.search(assistant_id="agent")

# Delete a specific job
await client.crons.delete(cron_job["cron_id"])

# Delete all jobs
for cron in crons:
    await client.crons.delete(cron["cron_id"])

⏱️ Cron Format

Standard cron format: minute hour day month weekday

• 0 9 * * * - 9:00 AM daily
• 30 14 * * 1 - 2:30 PM every Monday
• 0 */4 * * * - Every 4 hours

💡 Example Use Cases

• executive-ai-assistant – An AI agent acting as an Executive Assistant
• social-media-agent – Generates Twitter and LinkedIn posts from a given URL with human-in-the-loop approval
• agents-from-scratch – Step-by-step tutorial for building an email agent with Gmail API, HITL, and memory
• ff-take-bot – A “Take Bot” agent that runs in the background to keep a fantasy football league active
• ambient-agent-101 – Learn LangGraph basics by building an ambient agent that manages Gmail
• reddit-radar – AI agent for detecting or analyzing Reddit posts
• daily-brew – Automates publishing daily reflection content to a Slack channel