Short answer: for coding, start with PR Reviewer if you review code and loops if you run coding agents. For research, Research Bot gives cited answers and Competitor Watching tells you only when a rival changes something. For productivity, Bodyguard triages requests without ever replying for you. Below are fifteen verified picks across the three jobs, with how we chose them, who built them and when we last checked each one.
How we chose these Bots
There is no public install count or rating for Grok Bots, so this list is not a popularity ranking. It is an editorial shortlist from the verified shared Bots on GrokShelf, judged on four things we can actually check.
First, the Bot must be live: its x.ai share page has to load, show the "Add to Grok Bot" button and name the Bot and its creator. We re-check every listing on a rolling schedule, and each pick below shows the date of its latest check. Second, a real person has to stand behind it, so every pick links to its creator's own launch post on X. Third, the job has to be specific: a Bot that says exactly what it does beats one that promises to do everything. Fourth, it has to say where it stops. We favour Bots that read before they write, draft before they send and ask before anything irreversible, which is also what xAI recommends in its own documentation on routines and approvals.
What we did not do: we have not run each Bot against private repositories or inboxes, and we do not publish output-quality scores. Treat these as well-built starting points, then test them on your own low-stakes work.
Quick picks
| Job | Pick | Why it stands out |
|---|---|---|
| Code review | PR Reviewer | Leads every review with risk and missing tests |
| Running coding agents | loops | Wraps agent work in testable goals and evidence |
| Sourced research | Research Bot | Every claim comes with a source |
| Competitor monitoring | Competitor Watching | Silent unless something material changes |
| Request triage | Bodyguard | Sorts asks into three piles and never replies for you |
| Executive assistant | Jess | Answers routine staff questions from your playbook |
Best Grok Bots for coding
PR Reviewer · by @mustafaergisi
Point it at a pull request and it opens with the uncomfortable part: what could break, what isn't tested and where the context is too thin to judge. That ordering is the whole point. A human reviewer's attention goes to the risky part of the diff instead of being spread evenly across every line. It suits small teams where one person reviews most changes.
loops · by @mattyp
If you already hand work to coding agents, loops sits one level above them. It gathers repository context, writes goal prompts that can be tested, launches the work, reviews the evidence and guides the merge. The useful habit it enforces is proof: work isn't done because an agent says so, it's done when the evidence checks out.
Cursor Agent · by @ryanthawks
A local-only Bot that drives the cursor-agent CLI on your own laptop for experiments, while cloud agents stay the default for GitHub pull requests. Its creator sums it up as local hands, cloud for the repo. It doesn't deploy, submit or send anything without your explicit approval, which makes it a safe first coding Bot.
1000x Product Engineer · by @TomZarebczan
An opinionated full-stack builder for web apps that need to work on phones and desktops: Convex on the back end, TanStack and React on the front. What sets it apart is taste. It aims for simple, intuitive interfaces rather than scaffolding that merely runs. Pick it if your stack matches; skip it if it doesn't.
AI Harness Assistant · by @gheeunit
A small, practical one. It keeps the AI coding tools you already use — Codex, Grok Build, OpenCode, Cursor and similar CLIs — updated across Windows, macOS and Linux on a schedule. Nobody writes a blog post about updating CLIs by hand, but everybody loses time to it.
Best Grok Bots for research
Research Bot · by @ArthurMacwaters
Deep research with receipts. Every claim comes with a source, and it flags anything it couldn't verify instead of presenting it as fact. That one rule makes the output usable in places where a confident but unsourced answer is worse than no answer.
Competitor Watching · by @scheemunai
Snapshots your company against three to eight competitors — pricing, changelogs, features, job posts and messaging — and compares the snapshots every week. The design choice we like is restraint: it stays quiet unless something material moves, so you get an alert, not a weekly wall of noise.
Scout · by @adamta
Competitive intelligence aimed at organic growth. It checks rival sites daily for copy and pricing changes and tracks their search rankings and visibility in AI answers, then recommends what deserves your attention. Pair it with Competitor Watching if you need both strategy-level and search-level coverage.
Lurk · by @tinkerersanky
A Reddit researcher that turns seed keywords into evidence. It extracts exact quotes describing real pain points, builds an ideas pack from them and can draft a landing-page prompt. It only reads; it never posts or comments, which is exactly what you want from a research Bot loose on a community site.
Frontier Model Watch · by @GuleidAmina
A morning briefing on frontier AI releases drawn only from official sources such as xAI, Anthropic, Google and OpenAI. It reports verifiable changes rather than rumours, which is rarer than it should be in AI news. Adjust its default schedule to your own time zone.
Best Grok Bots for productivity
Bodyguard · by @liam_fallen
Protects your time from requests that probably don't need you. Incoming asks land in three piles: worth your time, deal with later and probably ignore. It never replies on your behalf and only talks to you, which makes it one of the easiest Bots on this list to trust with an inbox.
Inbot · by @matt_silberman
Inbox zero for everything that lands on you, not just email. During setup it asks you to connect each inbox you really use — email, Slack or Teams, calendars, Notion, messengers — then processes them on one repeatable routine, so incoming work gets handled the same way every time.
Jess · by @LoganARobison
A weekday executive assistant that recaps email, calendar, Notion and Slack, answers staff questions from a written playbook and posts client-call follow-ups. The playbook is the clever part: the team gets consistent answers without pulling you into every routine question.
Morpheus · by @liam_fallen
Decision support that actually decides. It narrows a choice to the strongest options, researches the facts that could change the answer and commits to one recommendation, naming the evidence that would reverse it. It's advisory by design: its configuration forbids acting on the decision itself.
Spark · by @vincentzhu
Not sure what to automate first? Spark asks a short set of questions, recommends or creates Bots that fit your answers and points you to the connectors that make them useful. It's a good first Bot because it helps you choose the second one.
What we left out, and why
Several popular Bots didn't make this list because they belong to other jobs — sales, recruiting and finance each deserve their own roundup. We also skipped Bots whose share pages didn't clearly state what they do on their own, and any Bot whose x.ai page had failed a recent check. If you built a Bot you think belongs here, submit it; verified Bots go live on GrokShelf straight away.
Looking specifically for an assistant or chief of staff? We cover that job in more depth in the best Grok Bots for a personal assistant. New to Grok Bots altogether? Start with what a Grok Bot is and how shared Bots work.
How to try one without regret
- Open the Bot's GrokShelf page and read what it does and where it stops to ask you.
- Add it from its x.ai share page, which gives you your own copy of its setup.
- Connect only the tools that job needs — a research Bot rarely needs your email.
- Run it once on low-stakes work and read the whole result.
- Keep sending, publishing and anything irreversible behind your approval.
Before adding any of these, read Using a Grok Bot safely.
References
- xAI: Introducing Grok Bot — launch announcement, August 11, 2026
- xAI docs: Skills, routines and automations — approval guidance and routine limits
- xAI docs: Shared Bots
- GrokShelf: all verified shared Bots
- GrokShelf: Using a Grok Bot safely