8 Grok Bot templates officially shared by SpaceXAI: turn workflows into reusable assistants

A template copies a way of working

Start with a writing assistant to understand templates. You want it to revise sentences according to familiar writing principles, remember your preferences, pull drafts from Notion, and drop the output into Typefully. If you only share the instruction "help me polish this," the recipient has to reconstruct all those requirements and connections on their own.

A template combines the relevant instructions, skills, memory, and official plugins. Skills define specific actions, memory holds reusable context or habits, and plugins let the assistant reach tools outside the conversation. Recipients use these configurations in their own environment, working with their own materials.

Palmer stresses in the accompanying tutorial that a template is a blueprint for rebuilding an assistant. The system selects content relevant to the workflow and excludes personal or internal information by default. What recipients get may still differ from the original assistant and may require additional setup. Custom code, scripts, and self-built MCP services aren't packaged automatically, and needed plugins have to be installed in the recipient's environment. MCP is essentially a way for AI to call external tools.

That distinction shapes how you should read the 8 examples that follow: you can reuse the author's approach to organizing work, but your own accounts, tools, and configuration still need to be connected.

Development: connecting task prep to code execution

Palmer's two development assistants work together. Loops handles preparation and organization for engineering tasks; Tech demos turns new technologies into reviewable demonstrations.

Loops: defining the task clearly first

Loops is Palmer's software engineering assistant. He discusses tasks with Loops first, and Loops then communicates with Cursor Agent, the assistant that actually executes coding work. Palmer calls Loops the "outer loop" assistant for engineering.

The inner and outer loops divide responsibilities: the inner code loop centers on writing, building, and testing; the outer loop also covers research, planning, forming a solution, deployment, and review. The original diagram keeps the two separate, a reminder that writing code is only part of the full engineering process.

When modifying a website, for example, Palmer has Loops read past contributions from several colleagues to understand existing practices. He also asks it to study competitors' product documentation, think through how similar problems might be handled, and consider how the current change should be verified. Loops organizes that background into a task prompt and then launches a cloud coding assistant. After that, Palmer moves into Cursor to continue.

He also saves the prompting patterns he reaches for repeatedly inside the assistant. Similar tasks can reuse the same approach of gathering context, forming a task, and iterating, so the next one doesn't start from a blank conversation.

Tech demos: turning saved items into something you can try

Each morning, Tech demos checks Palmer's X bookmarks, picks a new tool, and sends back a draft task prompt. That prompt follows the method he refined with Loops, then hands the development work to a coding assistant.

Palmer has explicit requirements for what comes back: screenshots and video. The demo can be deployed to Cloudflare, so a user can either take a quick look in Grok Bot, open the browser for hands-on interaction, or jump into Cursor to keep editing code.

The rotating-earth demo below is the visible result of that actual work. The video and screenshots come from the environment of Cursor Agent, the assistant writing the code. Grok Bot can communicate with that assistant, use its computer, and bring the results back.

Bookmarks marked "look at this later" now have a clear path forward: pick a tool, form a task, build a demo, check the result, and iterate. Palmer says he usually doesn't publish these demos, but having seen and tried them himself makes him more aware of what libraries are available and more likely to adopt or recommend them in projects.

Information management: making existing material serve the next decision

Development assistants turn materials into artifacts. Two other assistants take scattered information and arrange it into something immediately usable as reference, cutting down on situations where you have the information but can't find it when you need it.

Personal CRM: finding local friends before you visit a city

Personal CRM is a contact management assistant. It organizes people who follow Palmer on X and whom he follows back and writes them into Notion, arranged by country, state, and city. Before traveling, he can filter for friends at the destination and then decide who to meet for coffee or a walk.

The accompanying walkthrough clarifies the process: the assistant reads locations from public profile bios, pulls information via APIs or by browsing pages, and then builds a Notion database. The table retains fields like bio, location, and a confidence score for the location judgment. Locations are inherently ambiguous, and the curated records need to preserve that uncertainty.

The hard part of this kind of work is matching people to information and keeping it maintained. Palmer hands both parts to the assistant and works with the organized results. The video below starts with the contact demo and shows the path from making a request on the phone to filtering Notion by city.

X algorithm: a reference for post scheduling

The X algorithm assistant combines the X algorithm repository with native X connections to help Palmer decide when to post, what format to use, and how to engage with others. It takes over the small recurring decisions that come up during content scheduling.

Palmer also keeps his priorities straight: good content matters more than small optimizations. He finds the assistant most useful when he's unsure how to arrange content. It can cut down on hesitation, but it can't remove the question of content quality.

Content production: building quality standards into recurring work

Once your work starts producing drafts, icons, and other content, templates can also preserve the standards you want the results to meet.

Writing Bot: keeping human writing, then handling cleanup and sentence fixes

Palmer's Writing Bot draws on the writing books he relies on, including Style: Lessons in Clarity and Grace and On Writing Well. He usually writes first, then has the assistant handle grammar and difficult sentences. Typefully and Notion connections let the assistant work directly with existing drafts.

This very introduction to the 8 templates used Writing Bot in its preparation: it first requested summaries from his other assistants, then placed a draft in Typefully where he could start writing. It handles sentences but can also gather information scattered across different assistants into a single starting point for writing.

This workflow keeps human narrative and editing in place, while the assistant takes on material gathering and language cleanup. What's worth copying from the template are the recurring writing standards and tool connections.

Image Generation: keeping one visual style across a set of assets

The image assistant addresses visual consistency across sets of icons and characters. It first uses skills to send standardized prompts to Grok Imagine, then calls fal's background-removal model, and Palmer applies a unified color palette afterward.

Each of the three steps serves a purpose: the prompt spec constrains what gets generated, background removal makes assets easier to use downstream, and the unified palette keeps multiple assets looking cohesive in one project. Saving the whole pipeline in an assistant means the next batch of graphics can follow the same path.

Daily actions: from reminders to real phone calls

Two more templates extend work into everyday phone tasks. They also show that a task can be narrow and still worth reusing, as long as the input, action, and result are clear enough.

Skippy: turning parking restrictions into timely reminders

Skippy uses public San Francisco data to look up parking restrictions for a given location. Palmer can save frequent spots and receive reminders on weekday mornings before restricted periods begin. In the screenshot, the assistant distinguishes street-cleaning schedules for either side of the road based on the exact intersection.

It takes over the recurring chore of looking up a database or website. Once location, query, and reminder are linked, this small task can be handled by an assistant on the phone. Palmer would like to go further and connect a parking app so it can pay for parking on his behalf, but that's still an idea for later.

Dial Bot: giving an assistant a phone line

Dial Bot connects to Bland's calling capability. Palmer created a voice profile using his own voice, set up MCP connections and a calling skill, and the assistant can now place calls for him. He describes uses like navigating phone menus and waiting on hold.

A short video shows a very concrete task: calling a business to ask about that day’s opening hours, repeating the answer back for confirmation after hearing it, and ending the call. It demonstrates the question-and-answer flow of a phone call; voice menus and long hold times are further uses Palmer mentions.

The calling ability comes from an external service, and anyone reusing it needs to account for the corresponding setup. Custom MCP connections in particular aren't bundled with the template, so whoever opens the template link still has to supply their own configuration.

Turning your own working method into a template

After all these examples, the most practical starting point is something you've already done more than once: you know what materials you need, what order the steps happen in, and what a finished result looks like. Teach that one thing to an assistant first, then share the mature method.

The process Palmer demonstrates for creating and using a template breaks down as follows:

Open the image to see the interface details.

For complex templates, clear setup and configuration instructions are essential. Palmer expects the assistant to think through what another person would need—settings, API keys, extra tools—to reproduce the workflow, then bake those notes into the template itself. Recipients should also read through Context and Integrations first, to confirm the material lines up with what they’re about to do.

What all 8 examples share is a simple idea: capture the repetitive parts of your work so you don’t have to restate them. Developers save how they prepare tasks and define done. Writers save their editorial standards. Image work saves the production pipeline. The contacts and parking assistants save how queries get run and results get organized. When picking a first template, start from a concrete prompt: “Next time I do this, what requirements do I not want to spell out again?”

Browse the bot marketplace for a starting point

Grok Bot has an official template hub called Bot Marketplace, which pulls together bots shared by teams and other creators in one place. You can browse by category—engineering, sales, marketing, design, personal tasks, and more—or search directly for a bot or creator by name.

The marketplace spans far more than these 8 examples. You’ll find assistants for preparing sales leads, coordinating interviews, organizing creative assets, or even caring for houseplants. Pick a template that resembles the task you have in mind, look at its stated purpose, skills, and connected tools, and then decide which approaches are worth keeping and which settings need to become your own.

Keep exploringGrok Bot template marketplace

Pick a category or search for bots and creators.

Browse the marketplace ↗