Give your AI Superpowers.

Why it works

More done. More control.

Each worker runs separately, with its own permissions, memory and model. You decide what it can access and what it remembers.

• Keep your conversation focused. Each run has its own working context. Your AI gets the result without carrying every intermediate step.
• Limit access to the job. Each worker has its own permissions. Choose the apps and tools it can use, and keep unrelated connections out of reach.
• Remember only what you approve. Each worker keeps its own rules and facts. Proposed memory becomes available to future runs only after you approve it.
• Choose the right model for each job. Use decision models for structured judgments and language models for instruction-led work. Choose the kind that fits the job.

WorkerKit is the managed execution layer for AI assistants and agents.

Ask for the job. Get the result.

1. 01

   Your AI’s request

   Turn today's meeting notes into tasks.

2. 02

   Worker runs separately

   Reads your meeting notes and creates the tasks, using only the access you granted.

3. 03

   Back in your conversation

   5 tasks created. 2 items need your decision.

Start here

Start with a job you want off your plate.

• Browse all workers Every public kit, searchable by job, app and model.
• Meeting Prep Assistant Builds a prep sheet for each meeting on your calendar: who, history, open items, and talking points.
• Meeting Action Item Collector Reads your meeting notes each evening and files every action item into your task board.
• Follow-Up Tracker Tracks every email you are waiting on a reply to, drafts a polite nudge when a thread goes quiet past your window, and keeps a live waiting-on-others board.
• Email Reply Needed Screen Reads a window of mail and says, message by message, whether the owner owes a reply and who owes the next move.
• SEO Watchtower Watches Search Console weekly for ranking drops, indexing problems, and easy CTR wins, and files the fixes.
• Sheets Lead Qualification Returns lead-fit classifications from spreadsheet rows against your ideal customer profile.
• Company Knowledge Concierge Answers questions from your Drive, SharePoint, and Slack history, with citations to the source.
• Web Research Relevance Filter Finds and reads web pages on your research topic and returns relevant sources with typed judgments.
• Sales Anomaly Alert Watches your Shopify store's daily sales, orders and average order value, and alerts you only when one of them moves enough to matter.
• AI workers for Personal Every kit for personal.
• AI workers for Sales Every kit for sales.
• AI workers for Ecommerce Every kit for ecommerce.
• AI workers for Marketing Every kit for marketing.
• AI workers for Support Every kit for support.
• AI workers for Ops Every kit for ops.
• AI workers for Security Every kit for security.
• AI workers for Finance Every kit for finance.
• AI workers for Data Every kit for data.
• AI workers for Community Every kit for community.
• AI workers for Customer success Every kit for customer success.
• AI workers for Founders Every kit for founders.
• AI workers for Engineering Every kit for engineering.
• AI workers for Assistant Every kit for assistant.
• AI workers for Recruiting Every kit for recruiting.
• AI workers for Content Every kit for content.
• AI workers for Product Every kit for product.
• All roles Every role hub.

Choose a kit. Connect your apps. Review access. Run or schedule.

Certified marks a kit WorkerKit publishes itself. The badge is set by WorkerKit, never by a publisher.

Decision model workers

Give your AI a clear next step using Jev.

Turn messages, leads, and research into structured decisions your agent can act on.

1. 01

   Your agent asks

   Find the messages that may need my reply.

2. 02

   Jev decides

   Screens message previews against your questions and rules.

3. 03

   Decisions return

   Typed answers give your agent a next step. Unclear judgments stay visible.

• Email Reply Needed Screen Reads a window of mail and says, message by message, whether the owner owes a reply and who owes the next move.
• Sheets Lead Qualification Returns lead-fit classifications from spreadsheet rows against your ideal customer profile.
• Web Research Relevance Filter Finds and reads web pages on your research topic and returns relevant sources with typed judgments.
• Browse decision model kits Workers that classify, score, select, and route.

Schedules and limits

AI workers run in the background. You stay in control.

• Set the schedule. Decide when recurring work should run, in your own time zone, or run a worker on demand.
• Set spending limits. Per-run and daily budgets control what each worker may spend on model tokens.
• Review every run. Every run leaves a receipt with its outcome, status, and settled cost. Review the transcript for language workers or structured results for decision workers.

Run independent jobs in parallel within your plan’s limits. Keep worker settings and approved memory when you switch between supported assistants.

Chain one job into another, coordinate several workers and answer a run that stops to ask a question: that is AI worker fleet orchestration, and the MCP server page shows how your assistant drives it.

Questions

Questions before you connect.

• Should I use a decision model or a language model? Decision model workers return typed judgments such as categories, scores, and selected references, and can route work through defined rules. Language model workers follow instructions to research, write, and complete tasks. Both use the access you grant. Your agent can run a decision worker to find what needs attention, then use the results to finish the job.
• What is a worker, and what is a kit? A worker is set up to carry out one particular job. A kit is a ready-made worker: its instructions or decision rules, app permissions, and setup, packaged so you can run it with your own accounts.
• How does WorkerKit work with my AI assistant? You connect through a supported integration: the MCP server for Claude, Cursor and other MCP clients, the REST API for platforms that speak HTTPS, and guides of their own for Grok Bot and Meta Muse. Your assistant can then request work from a configured worker and receive its result. Each assistant in the row under the headline links to its own connection guide.
• Where does the work run, and does my computer need to be on? WorkerKit hosts the worker runs, so nothing of yours has to stay on. A scheduled worker keeps running when your laptop is closed, when your assistant is not open, and while you are asleep or away. Each run has its own working context, the result returns to the connected assistant, and the run receipt keeps the transcript or structured decision results.
• Which apps and data can a worker access? A worker uses only the connections and permissions granted to it. Review those permissions before running it. The result that returns to your conversation is a short digest, and everything else a run did stays on its receipt. That keeps your conversation short; it is not a filter on what a result may contain, since a digest can include information from the apps the worker read.
• What does a worker remember? Each worker has its own rules and facts that can carry across runs. Proposed memory must be approved by you before a future run can use it. Its temporary working context is separate from that approved memory.
• How do I control the model, spending and changes? For language workers, choose from supported models and providers, and use your own provider key where supported. Decision workers use their configured decision model and rules. Per-run and daily budgets cap what a worker may spend, and model tokens are billed at the provider list price. A running job can be stopped. Instructions are versioned, so an earlier version can be restored as a new one, although restoring does not undo an email, an app update or any other external action a run already completed.
• Can workers run automatically, and where do the results go? Workers can run on the schedule you set, in your own time zone, or on demand when you or your assistant ask. A scheduled run does not need you there to collect it: its report can be delivered to email, Slack, Microsoft Teams, Telegram, Discord or Notion, for each channel your account has connected. Every run also leaves a receipt you can open later, and you can choose whether a worker reports on every run or only when something failed or needs your decision.
• Do I need to build a worker, or know how to prompt or code? No to both. 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. If you want a custom job, the Kit Creator Studio is no-code as well, and the developer documentation covers the API.
• How much does WorkerKit cost? Browsing the directory is free and needs no account. The Free plan runs 5 workers with no card and no expiry, Pro is $29 a month for one person, and Team is $299 a month flat for up to 10 people. Model tokens are billed separately at the provider list price with no markup, paid from a prepaid wallet or billed to your own provider key.

What should your AI handle next?

• Find a worker Every public kit in the directory.
• Developer API The REST surface behind the MCP server.
• MCP server The generic connection guide.
• CLI Everything the fleet does, as commands.
• Create a worker The Kit Creator Studio.
• Grok Bot How Grok Bot connects to your AI workers.
• Muse How Muse connects to your AI workers.
• Instinct How Instinct connects to your AI workers.
• Hermes How Hermes connects to your AI workers.
• OpenClaw How OpenClaw connects to your AI workers.
• Claude How Claude connects to your AI workers.
• ChatGPT How ChatGPT connects to your AI workers.
• Gemini How Gemini connects to your AI workers.
• Microsoft Copilot How Microsoft Copilot connects to your AI workers.

Start with a ready-made worker for a job you do every day.