New and noteworthy

Agent Watch

Find new Bots and ways to use AI. Browse fresh listings from around the web, then see which ideas we have looked into more closely.

September 29 · OpenAI Dots

Another way to put an agent to work

Try sorting meeting follow-ups with a personal Dot. Specialist Dots are still limited to enterprise pilots. This walkthrough uses manual setup and has not been tested by Bot Cabinet.

Try a Cabinet job with a Dot
Bot directories and repositories

Current listings

Browse Bots and ways to use them. Choose All for tools, repositories and releases. Bot Cabinet has not tested these listings.

Checking directories…
Try an idea

Pick a first task

Try Daily Newspaper, which has one recorded Hermes test, or use the checklist to fix a bad AI answer. The checklist has not been task-tested.

Closer look

Ideas to try

Find the original source, setup requirements, and test status for each Bot or idea.

Bots

Worth trying

Bots for specific jobs, with source links and test status.

Listing reviewedGrok Bot

Import Bot

The creator says it copies selected agent instructions into Grok Bot, removes secrets and leaves the original setup unchanged.

Creator
Shub Gaur
Why it stands out
You could reuse an existing Bot's instructions when changing apps, rather than write them again. Its account connections and schedules may need new setup.
Accounts, tools and actions
The selected source setup and a Grok Bot account. The exact account and file access should be checked before use. Creates or updates Bots in the destination service after reviewing the imported setup.
Bot Cabinet plans
Test an adaptationTest it against Bot Porter's transfer checklist, including which instructions, tools and schedules need new setup. Similar Cabinet Bot: Portable Bot Packs
Limits and unknowns

Bot Cabinet read the marketplace description but has not installed or run Import Bot. Supported fields and secret removal have not been tested for each source app.

Listing reviewedMuse at Work

Diagnose My Agent's Mistake

Finds a likely cause for a bad AI result and proposes one concrete fix.

Creator
Creator not listed
Why it stands out
It helps you correct one bad answer and check whether the same mistake happens again.
Accounts, tools and actions
The failed prompt, the result, and any details needed to understand what went wrong. None described. The workflow returns a diagnosis and proposed correction.
Bot Cabinet plans
Improve an existing BotBot Cabinet prepared an Ops troubleshooting Skill for this method. It is ready for a manual test but has not yet been task-tested. Similar Cabinet Bot: Ops
Limits and unknowns

Bot Cabinet inspected the public Muse at Work listing but has not run the workflow. The listing does not identify a creator or show how well the diagnosis holds up across different kinds of failure.

Source reviewedWatching

Nasiko puts limits and records around a fleet of agents

Nasiko is an open-source control panel for agents that use the A2A standard. Its repository documents shared tool access, hidden provider keys, per-agent permissions, call tracing, cost records and limits that stop repeated or runaway work.

Why it matters
As people run more agents, the hard problem becomes control: what each agent may use, how much it may spend, what happened during a run and how to stop work that is going in circles.
Try this
If you run several agents, write down a spending ceiling, a time limit, a maximum number of handoffs and the tools each agent may use. Keep a run log and make sure one person can stop the work. Treat Nasiko as a developer tool until you can test it safely in a disposable setup.
Limits and unknowns

Bot Cabinet reviewed the repository but has not installed or run Nasiko. It requires Docker and several supporting services. Setup, upgrades and daily operation have not been tested here.

Updated 2026-09-24
Maker’s claimWatching

Grok Bot is pushing beyond one assistant into managed Bot fleets

An X post points to a 57-minute workshop in which SpaceXAI engineer Lauren Tan describes running many engineering Bots with a Chief of Staff Bot above them. xAI's own launch and engineering guides separately confirm that teams run multiple Bots in parallel and sometimes place one Bot in charge of coordination.

Why it matters
Separate Bots can build a change, test it and review the result. They need a shared task record and a clear owner for each next step.
Try this
Try one small software task with a coordinator, builder and reviewer. Save the commands, test results and handoff notes. Review the finished change before adding more Bots.
Limits and unknowns

The X post's claims about 20-plus Bots, 85 percent of engineers and overnight output have not been independently verified. xAI documents parallel Bot work, but that does not establish the post's numbers or the quality of the resulting code.

Updated 2026-09-24
Directories, guides and releasesBrowse the sources (24)
View sources

Current listings come from fourteen sources, including Grok Bot Marketplace, independent Bot directories and new agent repositories. You can also open the other sources directly; some have no dependable public feed.

Official marketplaces

Bot directories

Repositories and releases

Agent products

Agent infrastructure

Test status

Source reviewed: we read the original public source.
Tried by Bot Cabinet: we ran the named test.
Maker’s claim: the creator describes the feature; Bot Cabinet has not confirmed it.