Independent directorynot affiliated with xAI
Catalog refreshed Nobody pays to be listed or ranked
9,075 templates from 4,179 creators, each showing what it connects to, whether it asks first and how often it runs. “Checked” means the link was reachable — not a safety rating.
Give it a codebase or a directory and it hands back one two-sided card describing what the project is built with. Every statement on it is read from the .faf file, and it refuses to…
Start with a survey rather than a swarm. It walks your setup - firewall, hypervisor, out-of-band access, the web and media boxes - and proposes where the assistant should live and ho…
Hand it a template you are about to add and it reads the recipe — instructions, requested skills, plugin asks — as material to weigh, not orders to obey. The verdict comes back as bl…
A machine running a fleet of agents fills up quietly, and the danger in clearing space is throwing away something another agent still depends on. This one walks the filesystem and re…
A thinking partner for debugging with one twist: it will happily ask clarifying questions all day, but it refuses to hand you a fix until you clear four gates that prove you actually…
Will not let a coding task begin loosely: the objective, the boundaries and the finish line get fixed before anything runs. Repository work then goes to Cursor Cloud Agents on your a…
A print-prep engineer for people who would rather catch problems on screen than on the printer. It walks your part through geometry checks, bed-fit review and hole sizing, shows prev…
Adds a second browser to your Grok Bot machine so automation is not stuck with Chrome alone. You get an ad blocker from the start, passwords kept in sync for every bot on that box, s…
Sits over your collection of bots as a quality-control desk: it reviews their instruction packs, checks the routines that repeat on timers, and probes the rails connecting one bot to…
An odd and useful relay: a Grok bot whose output is meant for another tool entirely. It returns tight PLAN or REVIEW blocks sized to paste into a Cursor cloud coding agent, with none…
Made for small bench-mounted robot arms of the kind hobbyists buy. You walk the arm through its route once to capture the waypoints, it runs the sequence back at low speed, and then…
A smart-home foreman for households running Home Assistant. Lights, climate, solar and energy numbers are all in reach: it changes settings live over MCP, reads out the current figur…
Instead of opening with colour and layout, it maps the path somebody takes through a service and pins down what each screen must do at every moment: at rest, focused, still loading,…
A triage seat for teams that already run TypeSafe Jev. Incoming items are sorted into a category, given a severity, and answered with a plain go or no-go, all through typed calls ins…
Runs browser work in a Chrome window you can watch, clicking through pages with typed actions rather than guessing from pixels. When a page offers nothing usable to click, it steps d…
A small, quick classification layer built on the Jev harness. It answers the cheap, constant questions in a pipeline: which branch an item goes down, how urgent it is, which label fi…
If you already run a fleet of bots that grade, rank or verify things, this one rebuilds each of them as a sibling that judges a whole batch in a single sweep. Long rubric chains coll…
Point it at a question about your error tracking and it actually goes looking, through the Sentry MCP connection, the command-line tool and the web dashboard, rather than guessing fr…
Describe what you are trying to build and it returns a short list of candidate public data services from Korea's government portals. Each pick is kept brief: why it fits the idea, wh…
Looks after a published map of Grok bots. Ask it to sweep the catalogues and it files whatever turns up into a wiki of entries it already knows, then writes down the method behind ea…
Built for anyone about to ship an app, a site or a paid tier. It walks the happy path and then the failure path on purpose, and checks data-access settings from open tables to creden…
Works like a senior lighting designer brought in on a production or a building fit-out. It picks fixtures, runs the photometric numbers, lays out DMX, sACN or Art-Net control, and dr…
Ask for a small tool and it builds an actual page — a table, a form, a checklist — on a machine of its own, reachable from a phone or laptop on your Tailscale network, not the open i…
It turns your idea into a written spec, then hands the build to cloud agents who work as pull requests you review before anything merges. It runs the builds and assembles the store m…
Point it at an audio or video file and it runs more than one local speech model across the same recording, then hands you every transcript side by side. Where the versions disagree i…
It walks an online store from landing page to checkout the way a shopper would, then grades the trip metric by metric, marking each one passed or failed with evidence attached. Probl…
Describe the automation you want and it writes the workflow definition, then pushes it at your instance to see whether it actually runs. Broken nodes and mismatched parameters get fi…
Over time a bot's notes and memory fill with instructions for jobs that no longer exist, and it keeps honouring them. Optima reads through that material, works out which files belong…
Connects a real telephone number to an assistant by taking the fiddly provisioning off your hands: it signs you up with the telephone provider and installs an Asterisk server so the…
Describe the bot you have in mind, however roughly it comes out, and it works the concept up into three finished fields you can drop straight into the setup form: a name, a title and…
Point it at a dedicated machine and it handles the setup end to end: installing the engine, verifying each piece, and bringing up the dashboard with the local transformer running. Mo…
When API checks are not enough, it exercises the product the way a person would: working the interface on actual hardware, walking the paths a customer would take, recording each fai…
Hand it a single design problem and it comes back with three separate takes on it, each built out as a polished working prototype rather than a sketch or a mood board. With every one…
It starts from what you are actually trying to learn from the test, then picks a fidelity to match instead of over-polishing. Alongside the prototype it writes a prototype.md brief t…
Hand it an event log and it reconstructs how the process actually unfolds, not how the procedure notes claim. Its built-in PIX engine tests each trace against the expected flow and s…
Paste in an App Store or TestFlight rejection and it works out which guideline the review actually tripped and what the reviewer most likely saw in your build. From there it ranks th…