Claude daily briefing: one morning summary of your calendar, inbox, Slack and tasks

A Claude daily briefing is a scheduled run that reads your calendar, inbox, Slack and task list before you start work and gives you one short summary: today's meetings with the context behind each, replies you owe, threads that mention you, and tasks due. The useful versions link every item to its source and suggest actions without taking them.

What should a Claude daily briefing include?

Fix the order, so you know where to look every morning:

1. Today's meetings, with context. Who is attending, the last thread with them, the doc in the invite, and what you promised last time.
2. Replies you owe. Emails and messages where someone is waiting on you, oldest first.
3. Mentions. Slack threads where you were tagged since yesterday, with one line on what is being asked.
4. Tasks due today or overdue. From Notion, Linear or wherever your list lives.
5. Proposed actions. A clash to decline, a reply to send. Proposed, never done.

Anthropic's own daily briefing use case uses the same shape: urgent items, "Slack threads where I'm mentioned," "threads I'm not in but should probably know about," and "tasks due this week and anything blocking them." What it adds is the cross-link: one item that ties a metric to the thread behind it. That is where a briefing earns its keep.

How do you set it up yourself with Claude Code or Codex?

1. Connect the sources. Calendar, email, Slack and your task tool, each as a connector or MCP server.
2. Write the briefing prompt. Order, length limit, and what counts as urgent for you.
3. Schedule it. Claude Code routines run in the cloud, with a minimum interval of one hour and times entered in your local zone. Codex scheduled tasks on the web can use "connected tools," and desktop tasks that work in a local project need the computer on.
4. Watch the usage. Routines "draw down subscription usage the same way interactive sessions do." A Reddit user setting up a daily news briefing found research mode "seemed to consume a lot of Claude Pro usage." Scope the reads tightly.
5. Keep actions out of the run. A routine can use connector writes without asking, so give it read access or route any action through an approval.

Teresa Torres describes the local version in How to Choose Which Tasks to Automate with AI: she types /today into Claude Code each morning and watches it pull tasks from Trello into a today.md. That works because she is at her desk. A briefing that runs at 07:45 needs to run somewhere that is awake at 07:45.

How do you start it in Kavela in about 10 minutes?

Gmail and Google Calendar connect to Kavela in one click; Slack and Notion connect when Kav asks. Proposed actions arrive as approvals, so the briefing never acts on its own. The setup you still do:

1. Paste the prompt below into Kav in your Kavela workspace.
2. Connect Google Calendar, Gmail, Slack and, if you use it, Notion.
3. Read the first briefing and tell Kav what to cut. Those edits become rules for later mornings.

On weekdays at 07:45 <your time zone>, send me my daily briefing. I should be able to read it in two minutes.

Order:
1. Today's meetings from my Google Calendar. For each: who is attending, the last email thread or Slack conversation with them, the doc linked in the invite, and anything I promised last time.
2. Emails in my Gmail still waiting for my reply, oldest first, one line each.
3. Slack threads in <channels> that mention me since yesterday, one line on what is being asked.
4. Tasks due today or overdue in <Notion database link or your task tool>.
5. Suggested actions, such as declining a clash or replying to a thread.
Link every item to its source. Skip any section that is empty instead of padding it. No general news.
Never send, accept, decline or reply on my behalf. Put each suggested action in my approvals and do it only after I approve.
Connect Google Calendar, Gmail, Slack and Notion when you need them.
Run it once now so I can check it, then set it as a weekday routine.

What goes wrong with automated daily briefings?

• Too long to read. A briefing over two minutes gets skimmed, then skipped. Keep the length limit in the prompt.
• Context without sources. "Maria is worried about the deal" with no link is a rumour. Link every line.
• Acting on its own. Declining a meeting automatically is how you miss the one that mattered. Suggest, then approve.

If replies owed are the biggest section, hand them to AI email triage for Gmail, and fold competitor moves in from a weekly competitor watch. For the bigger habit shift, read how to use AI more efficiently.

Stuck on setup? Ask in the Kavela Discord.