From 15 coding agents by hand to a fleet of 200+, run by Grok Bots
In SpaceXAI's own guide, engineer Lingxi Li's bots start Cursor cloud agents, check every PR every 30 minutes and merge the low-risk ones. A teammate shipped 2,000+ PRs in a month, they report.
Done withGrok BotCursorSpotted· by SpaceXAI's Lingxi Li· Oct 18, 2026
Result Low-risk PRs merge themselves
“Before Grok Bot, I could manually manage 15 cloud agents at a time. Now my fleet manages more than 200 simultaneously, and we can scale further if we need to.”
Spotted: found in the wild. Details are as reported by the source above and have not been independently verified.
What they did
They have five engineer bots, each owning one area (iOS, desktop, Android, infrastructure, the harness), plus an operations bot, Jenny, who holds a 5 a.m. 1:1 with each of them. The bots start Cursor cloud agents with their skills and a detailed prompt, check the screenshots and proof that come back, and every 30 minutes review a shared Notion database of PRs for failing CI, merge conflicts and security findings. Highly confident, low-risk PRs merge automatically; the rest wait for their review.
How it went
They say they went from managing 15 cloud agents at a time by hand to a fleet that manages more than 200 at once, and that almost every morning there are tasks ready to merge. The guide also reports that a teammate shipped 2,000+ PRs in a month. These figures are SpaceXAI’s own, about its own team.
Try it yourself with Grok Bot
You are my engineer bot for [area of the codebase]. When I or Slack give you a task, start a coding agent with [my skills] and a detailed prompt that says what proof is expected (tests, before and after screenshots). Every 30 minutes, check the shared tracker of open PRs for failing CI, merge conflicts and review findings, follow up on each one, and mark a PR ready for review only when the proof matches the ask. Never merge anything touching [high-risk areas] without me.
A starting prompt written by Agents in the Wild. Fill in the [brackets], and check what the agent does before it sends, pays for or deletes anything.
A data scientist tried Cursor on a real project. Prototyping, code completion and commenting went well, but pandas code kept failing and it deleted a bunch of data files.
A 42k-line site with packages to upgrade: 19 easy, 42 medium and 14 hard, by Cursor's own sort. Cursor cleared the easy and medium ones, then a background agent moved it from Remix v2 to React Router v7.
A MongoDB-like database for a hobby project, written spec-first by Claude, Kiro, Qwen Coder and Cursor. They read zero lines of the code and checked it through tests. The first version took about two weeks, part time.
A Linux box mattered but wasn't their day job, and fixing it when it broke was anxiety-inducing. They gave Grok Bot root SSH access and their old setup notes, and now run nginx restarts and firewall rules from a phone app.
A messy, roughly 900-line Blender script, abandoned. Cursor turned it into a proper repo for all their 3D designs, then built shapes they described in plain language. "These aren't 1-prompt creations", they add.
A huge, convoluted codebase they'd taken over. One prompt in Cursor's Plan mode found the bugs and planned the fix, then the agent made it, where the client's previous team hadn't.