How to use ChatGPT tasks for dev chores, from briefs to pull requests

Here is how to use ChatGPT tasks: describe the job and its timing in a chat, confirm what ChatGPT proposes, and turn on push or email under Settings so the results reach you. For developers the two useful patterns are a timed task for a recurring brief and, on paid plans, a task that fires on GitHub pull request activity; for watching public changelogs and release pages, AyeWatch does the job better.

Everything below reflects OpenAI's help page on scheduled tasks in ChatGPT as of November 2026.

Pattern 1: a timed task

1. Open a chat and write the instruction with the schedule inside it: "Every Monday at 9am, summarize what changed in the Node.js and TypeScript ecosystems last week, with links."
2. ChatGPT answers with a confirmation card that shows the task's name and timing.
3. Go to Settings, then Notifications, and pick Push, Email or both. Phone push works once you create a task in the mobile app and grant permission.
4. Open the Scheduled page in the sidebar to see the next run, or to pause, resume, edit or delete the task.

On Free, a task runs once or at most daily, in a window like morning or night, and you can keep three active. Eligible paid plans add exact times and hourly repeats.

Pattern 2: a task that reacts to a pull request

Event-triggered tasks run in Work and are open to eligible Plus, Pro, Business, Enterprise and Edu accounts, not Free or Go.

1. In Settings, open Apps and connect your GitHub account.
2. Open Work and describe the event and what to do: "When a pull request in this repo is opened against main, list the files touched and flag any change to the migrations folder."
3. Check the Trigger, Condition and Prompt that ChatGPT drafts, and finish any authorization it asks for.
4. Manage the task from Scheduled like any other.

GitHub triggers cover supported pull request activity in an authorized github.com repository. Across all your event-triggered tasks, OpenAI allows up to 30 runs an hour and 720 a day, and it may group several events into one run. If an action needs your approval, the task waits until you give it.

Prompt tips that save runs

• Put the stop rule in the prompt. Monitoring tasks can end when a condition you set is met, such as "stop once version 5 ships".
• Ask for silence when nothing is new. A completed run can still use your plan's allowance, so do not schedule hourly checks you do not need.
• Keep tasks out of locked chats, GPTs and voice chats, where they are not supported. A task created in a project cannot read the project's files.
• Codex has its own automations. Scheduled tasks are the ChatGPT side.

Pattern 3: public release notes, with AyeWatch

A lot of developer watching is about pages you do not own: a framework's changelog, a cloud provider's pricing page, a deprecation notice, a competitor's docs. A ChatGPT task can search for news about them, but it is a prompt on a timer with a cap on how many you can keep active.

AyeWatch treats each of those as a monitor. Paste the URL, say what counts ("tell me when a new major version is announced"), and it reads the page for meaning and sends a short summary with sources when that happens. It can also follow topics across the web and public accounts on X, Reddit and YouTube.

The developer parts are on paid plans: email, Slack, Discord and webhooks for delivery, a REST API to create, list, update and delete topics, signed topic.update webhook events with AI summaries, and an MCP server so an assistant that speaks MCP can set up or pause monitors for you. Paid plans start at $9 a month, and the free AyeWatch preview needs no card.

Which to reach for

Use a ChatGPT task when the work lives in your own accounts, such as your repo, inbox or Slack, or when you want a written brief on a schedule. Use AyeWatch when the thing to watch is a public page, topic or account, and you want the alert on your phone, in Slack or at a webhook.