Grok Bots Directory
Submit a Grok Bot

Build a daily briefing Grok Bot: pick sources, schedule it, dedupe, keep it quiet

A morning briefing earns its place by being short, sourced and silent when there is nothing new. This guide picks sources, schedules a routine, fixes the format, adds a dedup ledger and quiets the bot, checked against x.ai documentation on 2026-10-31.

Checked on

The playbook

What you will get

A Grok Bot that sends one short briefing at a fixed time, built from sources you chose, without repeating yesterday's items and without messaging you on days with nothing to say. If you would rather pick an existing template than build one, read Best Grok Bots for daily briefings first.

Prerequisites

As of 2026-10-31, per get started: a plan that includes Grok Bot (Premium, Super or Ultra on The X Subscription, a linked SuperGrok or X Premium+ subscription, or a Grok Business seat), and the desktop or phone app signed in. Accounts on Privacy Mode (Legacy) are asked to switch to Privacy Mode first. Plans change, so check the page before paying.

You also need to decide where the briefing should be read: in the Bot's chat on your phone, or on desktop. Per Mobile apps, alerts for Bots you own go to the desktop app while it is in use.

Routine limits, as of 2026-10-31

From Skills, routines, and automations:

  • Up to 50 routines per Bot.
  • Schedules must be at least five minutes apart.
  • All your routines together can run 300 times an hour; each routine can run 200 times an hour. Runs past a limit fail with "Automation run rate limit exceeded".
  • The app keeps the 20 most recent run records for each routine.
  • A new routine waits for its next scheduled time; it does not run when created.
  • A test performs real work and can call connected tools.
  • Each run spends usage. The docs warn that "An hourly schedule, a short interval, or a Slack trigger on a busy channel can use a week of usage in a day."

One briefing a day is a trivially small load against these limits. The limit that matters is usage, not the run count.

What we could not verify

  • Web news access. Computer and apps does not say whether a Bot can search the general web. It documents a cloud computer with a browser, and read-only search of public X content without connecting an account. So "your Bot reads these news sites" is untested; try one run and see what the Bot says it could reach.
  • Importing a template. We did not import any template.
  • Any template's output quality. We read the skills and routines in our directory and ran none.
  • Controlling push versus silent delivery per routine. The docs we fetched describe a per-Bot Notifications switch only. They do not describe a per-routine setting to suppress output; the docs suggest putting boundaries in the instruction itself, which is what step 8 does.

Troubleshooting

Nothing arrives. Open the Bot's Details and check the routine's switch, "When to run" and your time zone under Settings, Bot, Timezone. A new routine waits for its next scheduled time. Imported routines are often paused or disabled; Intel Scout Bot ships its routine disabled.

It arrives at the wrong hour. The time zone is wrong. The Chinese-language template 左左的AI情报员 is written for 08:30 Asia/Shanghai on weekdays, so change the schedule if you are elsewhere.

The same story appears every day. Add a seen-items ledger and a dedup rule (steps 6 and 7). Ask the Bot to show you the ledger file.

You get "nothing to report" messages. Add the sentence "If this run finds nothing new, end the turn with no message" to the routine.

Runs fail with "Automation run rate limit exceeded". You have too many routines firing or a trigger that is too broad. Keep to one scheduled briefing.

Usage is climbing. Check for hourly or 15-minute routines, which came with some imported templates.

Sources

Checked on 2026-10-31 against get started, Skills, routines, and automations, Computer and apps, Mobile apps and Security. Re-check limits and menu paths after 2026-11-30, or sooner if xAI announces changes. Template behaviour is taken from the records in our directory, not from running them.

Build it in 11 steps

  1. Write down the reader and the one question

    Decide who reads it and what they should know within two minutes: market moves, AI launches, family logistics, or what happened on X. One briefing, one purpose. Briefings that mix six things get ignored. Planning step.

  2. Pick a small set of sources

    Choose five to ten sources and write them down. Good mixes: a few official blogs, a few reporters or accounts, and your own mail or calendar if relevant. Count what you can actually reach: connected apps are the reliable source. For public X content, Computer and apps says Bots can search and read it without a connected X account, read-only; the X plugin is for DMs, bookmarks and your home timeline and needs X API credits. For ordinary web sites the docs do not promise search, so test one run. 左左的AI情报员 fixes a list of 46 sources and AI Daily Radar tailors to your stack; both show how narrow a source list can be.

  3. Start from a template or from a blank Bot

    In the app open Connect apps, then Bot templates, then Import Bot (see Work with Grok Bot), or tell a fresh Bot in chat: "You are my morning briefing. Your job is X. Remember this." Read any imported routine before enabling it. Untested: we did not import a template.

  4. Connect only the apps the briefing reads

    Connect apps, search, add, finish the provider login, confirm the app appears under Installed. Apps can only do what your own account can do. Skip anything the briefing does not need; a read-only briefing should not hold send permissions. Source: Computer and apps.

  5. Decide the format

    Fix the shape once and write it into the instruction. Three formats work well: at most seven items, as AI Daily Radar does with a Friday wrap; a source, why-it-matters and what-to-do line per item, as Intel Scout Bot does; or a one-page paper with the schedule first, as Tradbot does for a household. Add a source link to each item, a word limit, and a rule to label unverified claims.

  6. Keep a seen-items ledger

    Tell the Bot to keep a short file of what it has already reported, keyed by link or title, and to read it before writing. Money Brief alerts once through a ledger; AI Daily Radar has a dedup skill that decides whether a repeat counts as an "Update:" or is skipped. Ask the Bot to show you the ledger after the first run, since file access is on the Bot's own computer.

  7. Write the dedup and update rule

    State it plainly: "Skip a story already in the ledger unless there is new information; if so, start the item with Update: and say what changed." Tell it to treat the same story from different outlets as one item and to prefer the primary source.

  8. Make it quiet

    Add one sentence to the routine: "If this run finds nothing new, end the turn with no message." Do not request a heartbeat. Keep to one scheduled briefing, since extra routines add usage. Tradbot's weekday digest is written to stay quiet when nothing needs you. The docs also say to put boundaries such as "Do not contact anyone" in the instruction itself.

  9. Set the time zone and schedule the routine

    Open Settings, Bot, Timezone, then ask the Bot: "Set up a routine to send my briefing every weekday at 07:30." Confirm the owning Bot, schedule, input source, expected result and what happens when a source is missing. Check When to run after saving. Limits are above. For breaking-news alerts, keep them separate and narrow, because a 15-minute check consumes usage fast; 时讯 pairs an hourly brief with a 15-minute breaking-news check, which is why many readers would keep only its nightly summary.

  10. Test once with Test

    Open the routine on desktop and choose Test; results appear in the Bot's chat. If Test is not offered, ask the Bot to run the briefing once in chat. Check that every item has a source, nothing is repeated, the length is right, and a second Test run reports nothing new.

  11. Run it for a week and edit

    Reply to the briefing in chat with corrections: too long, wrong sources, irrelevant item. Ask the Bot to remember the correction. Pause rather than delete while tuning, because deletion cannot be undone. Review sources monthly, as AI Daily Radar's monthly tune-up routine does.

Inspired templates6

Templates cited in this guide. Each page lists its skills, routines and required plugins; the collection compares them.

AI Daily Radar

AI Daily Radar

by Michael

Your morning AI brief, tuned to your stack. Launches, price changes, and benchmark claims from the last 24h, each tagged vendor-claimed or independently tested…

  • Memories10
  • Skills4
  • Routines2
左左的AI情报员

左左的AI情报员

by Bbbb

左左的 AI 情报岗：按固定 46 源做核验、去重与重要性评分，输出简洁中文情报日报；不凑数、不外发。含工作日晨报例程配方。

  • Memories5
  • Skills1
  • Routines1
Money Brief

Money Brief

by Oliver

Never get surprised by your own money: every weekday it tells you what is due, what posted and what looks wrong, then sends a weekly spending report and a mont…

  • Memories5
  • Skills2
  • Routines6
时讯

时讯

by wifibaby4u

每小时推送全球与中国大事件简报，全天盯重大突发，每晚发当日汇总，每条都附来源链接。适合想紧跟时事和市场的人。

  • Memories0
  • Skills1
  • Routines3

More playbooks

All playbooks