# About WorkerKit

> WorkerKit is a platform for AI workers: agents set up to do one specific job unattended, each packaged as a portable kit (the instructions, app permissions, memories, schedule and more) that you can have up and running in about a minute with no code. Workers are model agnostic across Claude, ChatGPT, Gemini and Grok with no markup on model tokens, they run secure and under your control, and a public kit marketplace is built in.

This document is the full, machine-readable product description for AI agents
and crawlers. It lives at https://workerkit.ai/about.md and is indexed from
https://workerkit.ai/llms.txt. The human site is https://workerkit.ai.
Last updated: 2026-08-04.

## What WorkerKit is

An AI worker is an AI agent set up to do one specific job on its own, on a
schedule or a trigger, using the apps you connect. Some teams call this an
agentic workflow. Because a worker runs unattended, it ships with the three
things an unattended job needs: a written instruction, explicit app
permissions, and rules for what it may read and touch.

A worker kit is a ready-made AI worker in a portable package: the
instructions, the app permissions, the memories, the schedule and everything
else one specific job needs. You pick the kit, connect your apps, and the
worker is on the job, about a minute from pick to first run. Taking a kit
mints a worker in your own account, with its own key and only the access you
granted.

The public kit marketplace is one of the platform's advantages rather than
the product itself: the directory at the site root is free and anonymous to
browse, every published kit gets a permanent page, and anyone can publish
with the no-code Kit Creator Studio, publicly or privately.

## Core features

- **Workers and kits.** The unit of the product is the worker: instructions,
  app permissions, memories, schedules, triggers and more, packaged as a
  portable kit anyone can run with their own accounts.
- **Up and running in about a minute.** Pick the kit, connect your apps, run.
  No prompt engineering, no glue code, no APIs to wire together.
- **Model agnostic.** Claude, ChatGPT, Gemini and Grok run the same kits, and
  every kit page shows how each model performs on it, so you can run the
  model that does the job best and switch the moment another one wins.
- **No markup on model tokens.** Tokens bill at the provider list price to
  the cent. WorkerKit's own fees are the plan price and the wallet top-up fee
  (5.5% with a $1 minimum), both stated in the open, and running workers
  through your own AI client is free on every plan.
- **Secure and in control.** Access is granted per app and per level
  (Off / Read / Write), can be narrowed or revoked at any time, and the
  safety layer (app firewall, redaction, contact rules) ships on every plan.
  Every run leaves a receipt.
- **A public kit marketplace built in.** Discover kits others published,
  publish your own with a permanent page, or keep kits private.

## How it works

Three moves take you from browsing to a working worker, in about a minute:

1. **Pick the job.** Browse kits by the work they do, not the tools they use.
   Every kit states its one job in a sentence and shows what it touches before
   you take it.
2. **Connect your apps.** Sign in to the accounts the kit asks for. You grant
   each one yourself, you see exactly what it can read or write, and you can
   switch any of it off later.
3. **Deploy and run.** Send the worker to the runtime you already use. It runs
   on your schedule or on demand, with your rules in force from the first run.

No prompt engineering, no glue code, no APIs to wire together.

## What every kit page tells you

Four things, before you take it:

- **The job**: what the worker does, in one sentence.
- **The loadout**: the apps it uses and the exact access each one needs.
- **The trigger**: whether it runs on demand, on a schedule, or reacts to
  events.
- **The proof**: real download and star counts, the publisher behind it,
  verification badges, and how each AI model scored on this kit.

## Access and control

A worker is only as trustworthy as the access behind it, so the access is
yours to give, to narrow, and to take back:

- **See it before you take it.** Every kit page lists the apps it uses and the
  access each one needs, in plain words, while you are still deciding.
- **Grant it app by app.** You connect each account yourself and choose what
  the worker may do there, per app and per level (Off / Read / Write). A kit
  cannot reach anything you did not hand it.
- **Take it back anytime.** Narrow a permission or switch the whole thing off
  later. The worker keeps running with whatever is left, and never quietly
  regains what you removed.

The safety layer (the app firewall, redaction, and contact rules) ships on
every plan including Free. It is never sold as an upgrade.

## Model choice

Claude, ChatGPT, Gemini and Grok run the same kits: a kit brings the
instructions, not the model.

- **Swap anytime**: change a worker's model whenever you like, or send a
  single run to a different model to compare them on the same job.
- **Graded per model**: every kit page scores itself per model, and each grade
  is marked with where it came from, measured from real runs or declared by
  the publisher.
- **No token markup**: model tokens on hosted runs bill at the provider's list
  price.

## Where workers run

Two lanes, and neither hides a token markup:

- **Bring your own runtime.** Point your own machine, assistant or MCP client
  at WorkerKit and it is free and unlimited on every plan, Free included. You
  pay your model provider directly and WorkerKit takes nothing from that bill.
- **Let WorkerKit run it.** The managed runtime hosts the run: workers follow
  wall-clock schedules in your own time zone, fire on webhook triggers, or run
  on demand, and every run leaves a receipt in the activity feed. Model tokens
  are billed at the provider list price with 0% markup, drawn from a prepaid
  wallet; new accounts start with trial credits. The only fee is on loading
  the wallet: 5.5% with a $1 minimum, so the smallest $5 load charges $6.

One-click deploy to external runtimes is in the works. The destinations shown
on the site today: Claude Cowork, Hermes, OpenClaw, Lovable and Replit among
agent runtimes, and Vercel, Heroku, Google Cloud, Microsoft Azure, Cloudflare
and AWS on the cloud and serverless side.

## Pricing

Plan of record (also at https://workerkit.ai/pricing):

- **Free, $0 forever.** One user, 5 workers, 1 email account. 500 tool calls a
  day, 60 an hour. Basic memory, 1 day of logs, install and publish kits
  including private kits, and the whole safety layer ungated. No card, no trial
  clock.
- **Pro, $29 per month ($290 per year).** One user, 50 workers, 5 email
  accounts. 5,000 tool calls a day. Advanced memory, 30 days of logs with
  alerts, IP allowlists and block rules, email support.
- **Team, $299 per month flat ($2,990 per year), up to 10 users, not per
  seat.** 500 workers, 10 email accounts, 50,000 tool calls a day pooled across
  the team. Private kits shared across the team, 365 days of logs, team
  dashboards, roles and permissions, central worker admin, one shared wallet
  with spend controls, priority support and an onboarding session.
- **Enterprise, custom pricing.** Custom limits and priority lanes, SSO and
  SAML, identity sync with offboarding, SIEM export, on-prem or EU
  zero-retention models, compliance pack (SOC 2, DPAs, audit support),
  dedicated TAM and SLA.

An email account is one Google or Microsoft account a worker signs in as: Gmail
or Outlook, with the calendar, files and tasks that come with it. A tool call is
one thing a worker does in a connected app: reading a thread, sending a reply,
adding a calendar event. Yearly billing works out at two months free. Plans change or cancel at any time; workers, kits and connected apps
stay where they are, only the limits move.

## Frequently asked questions

**What is an AI worker?**
An AI worker is an AI agent set up to do one specific job on its own, on a schedule or a trigger, using the apps you connect. Some teams call this an agentic workflow. Because a worker runs unattended, it ships with the three things an unattended job needs: a written instruction, explicit app permissions, and rules for what it may read and touch.

**What is a worker kit?**
A worker kit is a ready-made AI worker: the instructions, the app permissions, the memories, and the schedule for one specific job, packaged so you can run it with your own accounts. You pick the kit, connect your apps, and the worker does the job it was written for.

**Do I need to know how to prompt or code?**
No. The hard part of an AI worker is the instruction, the access rules, and the wiring, and a kit ships with all three already written and tested by its publisher. You connect the accounts it asks for and run it.

**What can a kit see in my accounts?**
Only what you grant it. Every kit states the apps it uses and the exact access it needs before you take it, you approve each connection yourself, and you can narrow or switch off any of it afterwards. A kit cannot reach an account you have not connected.

**Where does a worker actually run?**
Taking a kit mints a worker in your own account, with its own key and only the access you granted, and there are two ways to run it. Point your own machine, assistant or MCP client at it, which is free and unlimited on every plan. Or let WorkerKit host the runs, so the worker follows its schedule unattended, with model tokens billed at the provider list price and no markup.

**Can a worker run on a schedule or react to events?**
Yes. A hosted worker runs on the schedule you set in your own time zone, fires on webhook triggers from the systems you already use, or runs on demand when you ask. Every run leaves a receipt in the activity feed showing what happened, so unattended never means unwatched.

**Which AI models do kits work with?**
Claude, ChatGPT, Gemini and Grok all drive the same kits. Every kit page grades itself per model, and each grade is marked with where it came from, measured from real runs or declared by the publisher, so you can weigh it before you deploy. Filter the directory by model to see only the kits that fit the one you already use.

**Can I switch a worker to a different model?**
Yes, at any time. A kit brings the instructions, not the model, so a worker is never tied to one provider. You pick the model when you deploy, switch it whenever you like, and can send a single run to a different model to compare them on the same job.

**How much does WorkerKit cost?**
Browsing the directory is free and needs no account. The Free plan runs 5 workers with 500 tool calls a day and the complete safety layer, with no card and no expiry. Pro is $29 a month for one person, Team is $299 a month flat for up to 10 people, and running workers through your own AI client stays free on every plan.

**Can I build and publish my own kit?**
Yes. The Kit Creator Studio is a no-code builder: describe the job, pin the apps and the access each one gets, write the instruction, then validate and publish. Published kits get their own page and permanent address in this directory, and a kit can also stay private instead of going public.

**What happens when a publisher updates a kit?**
A kit keeps its address for life, so an update lands on the same page rather than a new one, and every link to it keeps working. Each kit page shows when it was first published and when it was last updated.

## Pricing questions

**Is the free plan actually free?**
Yes. No card, no trial clock, no expiry. Free gives you 5 workers, 1 email account, 500 tool calls a day and the complete safety layer, and it stays that way for as long as you use it. You move to a paid plan when you want more scale, not when a timer runs out.

**What counts as a tool call?**
A tool call is one thing a worker does in a connected app: reading a thread, sending a reply, adding a calendar event. Free allows 500 a day and 60 an hour, Pro allows 5,000 a day, and Team allows 50,000 a day pooled across everyone on the account. The counter resets daily.

**What happens when I hit the daily cap?**
Nothing breaks and you are never locked out. On Free the workers keep running at free rates until the counter resets. On a paid plan you can add $5 packs, and once those are used the workers fall back to free rates rather than stopping. We do not hold your running jobs hostage to an upgrade.

**Do you mark up model tokens?**
No. Model tokens on the hosted runtime are billed at the provider list price with 0% markup, to the cent. The only fee is on loading the wallet: 5.5% with a $1 minimum. Loading the $5 minimum costs $6 and buys $5 of tokens at list price.

**Can I avoid the wallet entirely?**
Yes. Point your own runtime at WorkerKit through the MCP service and it is free and unlimited on every plan, including Free. You pay your model provider directly and WorkerKit takes nothing from that.

**Is the firewall a paid feature?**
No, and it never will be. The app firewall, redaction and contact rules ship on every plan including Free. A worker that runs unattended needs those controls to be safe to run at all, so selling them back to you would mean selling an unsafe free plan.

**How does Team billing work?**
Team is $299 a month flat for up to 10 users, not per seat, so a tenth teammate costs nothing extra. Free and Pro are single-user accounts, so Team is the plan where more than one person shares a fleet. It carries one shared wallet with spend controls, roles and permissions across the fleet, and central worker admin including member on and offboarding.

**Can I switch plans or cancel?**
Yes, at any time. Monthly plans cancel at the end of the period. Yearly is billed once and works out at two months free. Your workers, kits and connected apps stay where they are when a plan changes; only the limits move.

**When do I need Enterprise?**
When workers stop being one person's tool and become something the company runs on. At that point they have to run under your own sign-in and on your own network: SSO and SAML, identity sync with an offboarding cascade, SIEM export, on-prem or zero-retention models, and a compliance pack with SOC 2 and DPAs. Pricing and limits are both set with you, around what you actually run.

## Site map

- Kit directory (site root): https://workerkit.ai/
- Kit pages: https://workerkit.ai/kit/{slug}
- Publisher profiles: https://workerkit.ai/publishers/{slug}
- Category hubs: https://workerkit.ai/for/{category}
- Pricing: https://workerkit.ai/pricing
- Kit Creator Studio (sign-in required): https://workerkit.ai/studio
- Terms of service: https://workerkit.ai/terms
- Privacy policy: https://workerkit.ai/privacy
- Acceptable use policy: https://workerkit.ai/aup
- Security: https://workerkit.ai/security
- Crawler index: https://workerkit.ai/llms.txt
- Robots: https://workerkit.ai/robots.txt

## Notes for AI agents

- Every page on workerkit.ai is client-side rendered: the raw HTML of any
  route is the app shell, and the rendered DOM carries the content. This
  document restates the stable facts so they are readable without executing
  JavaScript.
- Rendered public pages carry JSON-LD: WebSite with SearchAction, ItemList and
  FAQPage on the directory; SoftwareApplication and BreadcrumbList on kit
  pages.
- Live catalog data (kit names, counts, stars, downloads) lives only on the
  rendered pages and is deliberately not restated here, so this document can
  never contradict the catalog.
- Kit, publisher and category URLs are stable, human-readable, server-minted
  slugs. A kit keeps its address for life.
- Crawling every public route is welcome; see https://workerkit.ai/robots.txt.