Automations in Grok

On July 16, 2026, xAI shipped Automations in Grok: saved jobs that Grok executes on a schedule or in response to an incoming email, then reports the result back. The feature is live on grok.com and in the Grok apps for iOS and Android, with a dedicated hub at grok.com/automations.

An automation is authored the same way you write any Grok message. You describe the work in natural language, optionally attach files, wire in connectors and skills with @ mentions, pick a mode, name it, and save. From that point forward, Grok treats every run as a fresh request against the same instructions with current data.

What shipped

Two trigger types are supported. Schedules can fire once, daily, on weekdays, weekly, monthly, or yearly, at a time you choose in your timezone. Email triggers watch the inbox and fire when an incoming message matches filters on sender, recipient, or subject, passing the email itself as context so Grok can reply to the actual message.

A Run now button lets you fire an automation manually, which xAI positions as the testing path. Every automation can be paused, resumed, edited, or deleted at any time, and each run opens a real conversation saved to run history that you can read in full or continue chatting inside. Reporting is configurable per automation: email, in-app notification, both, or neither.

Automations can also be created directly from a chat prompt (xAI’s example: “check the news every morning and flag anything about pricing”), and the Automations page ships with suggested templates.

1. 1Describe the job in natural language
2. 2Attach files, connectors, and skills with @ mentions
3. 3Pick a schedule or email trigger
4. 4Test with Run now, then save

On availability: scheduled automations are available to everyone. Email triggers are included with SuperGrok.

Key details

• Launch date: July 16, 2026.
• Surfaces: grok.com and Grok apps for iOS and Android. Hub at grok.com/automations.
• Schedule frequencies: once, daily, weekdays, weekly, monthly, yearly. Time is user-selected in the user’s timezone.
• Email trigger filters: sender, recipient, subject. The matching email becomes run context, and Grok can reply to the message itself.
• Connectors and skills: attached at authoring time via the @ mention pattern.
• Manual execution: Run now button.
• Run artifact: every run opens a real conversation saved to run history. Threads are readable in full and resumable.
• Reporting channels per automation: email, app notification, both, or none.
• Lifecycle controls: pause, resume, edit, delete.
• Authoring paths: Automations page, templates, or from chat by asking Grok to set one up.
• Tier gating: scheduled automations available to everyone. Email triggers included with SuperGrok.

Why it matters

Grok is late to scheduled agents, but it enters with the two primitives that actually matter: a time-based trigger and an event-based trigger tied to a real data source the user already has. The bet here is not novelty. Every major assistant is converging on the same shape, a saved job with a schedule, a set of tools, and a report-back channel. What differs is which triggers each platform exposes, how deeply the assistant’s own connector graph is wired in, and whether the run leaves behind an inspectable artifact.

For a cross-platform audience deciding where to place agent bets, three things in this release are worth naming.

1. Email as a trigger is the highest-value ambient signal most knowledge workers have. Platforms that treat it as first-class turn the assistant into an inbox agent without the user standing up Zapier or a custom webhook.
2. Every run is a full conversation, not a black-box job log. That matters for debugging, for handoff, and for the moment an automation gets something wrong and the user needs to correct it in place.
3. The create-from-chat path collapses the authoring surface. There is no separate builder to learn, the same design choice pushing adoption of assistant-native workflow features across the category.

Scheduled automations are free; email triggers are gated to SuperGrok. That split signals xAI reads inbox-triggered agents, not cron jobs, as the paid behavior worth locking behind the wall.

How it compares

Without a cross-platform snapshot in this brief, the honest positioning is qualitative. Scheduled tasks and event-driven agent runs are not new as a category. What varies across assistants is the depth of native trigger surface (schedule only vs. schedule plus email plus webhook plus app events), whether runs are inspectable as full conversations or opaque job logs, and whether the connector graph the assistant already uses in chat is the same one available inside a scheduled run.

Grok’s release lands on the more capable end of the schedule-only baseline that has been the default across the category, because it ships email triggers alongside schedules at launch and treats the triggering email as first-class context. The “every run is a real conversation, resumable in place” design is the other differentiator worth flagging. It is the opposite of the batch-job pattern most automation tools default to.

What we cannot compare from the source alone: latency of run execution, concurrency limits, how many automations a user can hold at once, and whether email-trigger inbox coverage extends beyond a specific provider set. Those are the axes on which cross-platform comparisons will actually be decided.

Who should care

Solo operators and founders running lean. The email-trigger path is the highest-leverage piece here for anyone who lives in their inbox: investor updates, customer escalations, vendor pricing changes, and job-alert-style watches can be handed to Grok with a filter and a set of instructions. If you already pay for SuperGrok, this is worth an hour today to migrate two or three recurring inbox chores.

Builders shipping agent side projects. Grok Automations is not a replacement for a proper workflow platform, but it is a fast way to prototype an “assistant that watches X and reports Y” loop without standing up hosting, a scheduler, or a webhook receiver. Use Run now to iterate on the instruction, then hand it off to schedule or email.

Growth, ops, and research teams. Morning briefs, competitor monitoring, pricing-change flags, and news roundups are the templated use cases xAI is pointing at, and the ones most teams will actually run. The value is in the run history being a readable conversation, so when a brief looks wrong you can inspect exactly what Grok saw.

Engineering leaders and platform owners. The signal to track is not the feature but the shape: schedule plus email plus resumable conversation plus connector graph. If your standardization bet is on an assistant that exposes only cron-style scheduling, Grok’s release is a data point that the bar is rising.

Investors and analysts watching the assistant-vs-workflow-tool boundary. This is another release compressing the space between “AI chat” and “workflow automation” into a single product surface. Track how quickly the rest of the field matches the email-trigger primitive.

What changed practically

Grok users now have a persistent job layer on top of chat. Concretely:

• Authoring surface. An automation is authored like a chat message. Attachments, connectors (referenced with @), skills, and modes are all in scope. Automations can be created from the Automations page, from a template, or by asking Grok to set one up mid-chat.
• Triggers. Schedule options are once, daily, weekdays, weekly, monthly, yearly, at a user-chosen time. Email triggers filter on sender, recipient, or subject, and when they fire, the matching email is passed in as context, with Grok able to respond to that message directly. Run now is available for manual execution and testing.
• Execution and inspection. Each run opens a real conversation saved to run history. Users can open any past run to read the thread end to end and continue from where Grok stopped. Notification behavior is per-automation: email, app notification, both, or none. Any automation can be paused, resumed, edited, or deleted.
• Availability. grok.com and iOS/Android. Scheduled automations for all users. Email triggers included with SuperGrok.

What to watch next

Three things are worth watching over the next quarter.

First, whether xAI extends the trigger set beyond schedule and email. The obvious next primitives are webhook triggers, calendar events, file-drop triggers on connected storage, and message triggers on chat platforms. Every one of those is table stakes for the automation platforms Grok is now brushing up against. Shipping any of them would move Grok from “assistant with schedules” toward “assistant as workflow runtime.” Expect at least one to land before this feature’s first anniversary.

Second, whether the “every run is a real conversation” design holds up under real usage. It is the right design for debugging and keeping the human in the loop, but it also means run history grows fast for high-frequency automations. Watch for retention policies, search over run history, and whether Grok adds a summary layer on top of the raw conversation log.

Third, the SuperGrok gating on email triggers is a soft price signal. If email triggers drive meaningful conversion, expect more inbox-adjacent capability on the paid tier: reply drafting inside the automation, multi-account inbox coverage, richer filter logic. If they don’t, expect the gating to soften.

The broader test is whether users who set up an automation on day one still have it running in ninety days. Scheduled-agent features have historically had a novelty problem: easy to create, easy to forget. The presence of pause/resume/edit in the lifecycle controls suggests xAI knows this, and retention is the metric that matters most for judging this release six months from now.

Open questions

• Which email providers are supported for triggers, and how is the inbox connected? The announcement names email as a trigger and describes filter fields but does not spell out provider coverage or auth model.
• What are the concurrency and quota limits? Nothing states how many automations a user can create, how frequently they can fire, or what happens when many email triggers match at once.
• How is failure surfaced? Runs are described as full conversations, but the announcement does not address retries, partial failures, or what appears in run history when a connector call errors.
• What is the SuperGrok scope for email triggers, exactly? “Included with SuperGrok” is stated, but per-user limits inside that tier are not detailed.
• Are there team or workspace semantics? The release reads as single-user. Shared automations, org-level ownership, and audit trails are not mentioned.
• Is there API or programmatic access to an automation’s runs? Whether that history is exportable or callable from outside Grok is unstated.
• What connectors are actually available inside an automation, and does that set match chat? The @ mention pattern is described, but the connector inventory is not enumerated.

Source: x.ai