How to Make Claude Notify You When a Task Is Done

Claude cannot send you a push notification the way a messaging app does. Whether you can get an alert when Claude finishes a task depends on which Claude surface you are using and whether you are willing to add a small script. On Claude Code, it is straightforward. On Claude.ai, you check back.

A reliable notification is only half the picture. You also need a clear record of what was done and what comes next. A tool like fenbs keeps that record in one place so you do not have to scroll through the conversation to find out where the task stands.

Claude Code: a notification script

Claude Code runs in a terminal and can execute shell commands as part of a task. That means Claude can run a notification command at the end of its work. On macOS, the say command reads text aloud. On most Linux desktops, notify-send shows a system notification. On Windows, PowerShell can display a toast notification.

Ask Claude at the start of the session: "When you have finished the task, run a notification so I know it is done." Claude can write and execute the appropriate command for your operating system. You can also include this as a standing instruction in your CLAUDE.md or system prompt so it applies automatically.

say "Task complete"

Claude.ai: check back manually

Claude.ai is a browser-based chat interface. It does not have a mechanism to ping you when a response is ready outside the browser tab. Your browser may show a favicon change or a tab title update when a response arrives, but there is no push notification to your phone or desktop.

For long-running Work tasks in ChatGPT or similar agentic products, the product's own interface handles status updates. Claude.ai does not currently have an equivalent background task feature that notifies you on completion.

If you want a notification on your phone when a long Claude.ai response finishes, one workaround is to keep the tab active on a desktop and use your phone to check the result directly. Browser extensions that watch for page changes can also serve as a proxy notification. Neither approach is as clean as a push alert, but they reduce the need to keep refreshing.

Checking the output, not just the notification

A notification tells you the task finished. It does not tell you whether it finished correctly. Treat the check as a separate step from the acknowledgment.

For a code task, run the tests after Claude Code signals completion. For a research summary, read through the findings before you pass them on. For a file operation, open the file and confirm the expected change is there. A notification that fires after an error is still a notification.

Some tasks are fast to verify; others take longer. Build the check into your workflow. If the output passes, update the task to done. If it does not, record what went wrong so the next run benefits from that information.

A quick review also gives you a chance to record any edge case Claude handled differently than expected. That note becomes a standing instruction for future sessions: "When the export file is missing column D, stop and ask before proceeding." Edge cases found and recorded in one session prevent the same stumble in the next, which saves more time than the notification itself.

Third-party automation

If you are using Claude via the API in your own workflow, you can trigger any notification mechanism after the API call completes: send a Slack message, post to a webhook, call a push notification service. The API response is the signal; what you do with it is up to your code.

What to record so the handoff is useful

A notification tells you the task is done. It does not tell you what was done, what changed, or what comes next. Record those facts while the session is fresh: what was completed, any file that was modified, any step that turned out differently than expected, and the next action.

That record is most useful when it lives outside the conversation. In fenbs, Where we left off captures exactly this: what the assistant finished, what is still in progress, and what is waiting on a person. The next session or the next person picks up from there.

How to do this with fenbs

1. Create a free account and add the task you want Claude to work on.
2. Connect Claude over MCP from Integrations and ask it to read the project context first.
3. If you are using Claude Code, ask it to send a notification when done using the appropriate command for your OS.
4. Before the session ends, ask Claude to save progress: what was completed, what is still open, and the next step.
5. Return to fenbs to read Where we left off and verify the task status before starting the next session.

For more on managing task handoffs across sessions, see what is Claude Cowork and what is Claude Code.