AI workers for Meta Muse.

Start here

Let Muse set itself up.

Muse calls outside services over their APIs and keeps the key in its own credential entry, so one prompt does the setup. Your part is a key and a paste.

1. 01

   Create a manager key

   At AI access, name it Muse, start with the read scopes and copy it once.

2. 02

   Paste the prompt into Muse

   Muse reads the setup guide and opens its own key entry for WorkerKit.

3. 03

   Enter the key where Muse asks

   In Muse's hosted key entry, never in the chat. Muse then checks what the key allows and lists your workers.

• Paste into Muse Set up WorkerKit for me, following https://workerkit.ai/muse/setup.md. Ask me for my WorkerKit manager key through your hosted API-key entry, never in this chat. Call https://api.workerkit.ai with it, looking up each operation in https://api.workerkit.ai/openapi/workers.json. Install the skill from https://workerkit.ai/muse/SKILL.md and follow its rules. Call key_info and tell me which account and scopes you have. List my workers and recent runs. If there are none, ask what job I want done and suggest a ready-made kit for it. Ask me before you run, install or change anything.

• Create a manager key At AI access, admin only. It is shown once, so keep it to hand.

The setup guide has every step, and the API page the contract behind it.

Why WorkerKit

Control your AI workforce from Muse.

Ready-made, self-learning AI workers with only the access each job needs, a budget you set, and a record of every job and action.

• Put AI workers on jobs with real ROI. Install a kit for work your team already repeats, and a worker arrives with its instructions, app permissions and schedule already written and reviewed. Start with the job that costs the most time, or build your own kit.
• Turn your AI workers’ jobs and data into shared company knowledge. Each job’s instructions, apps and limits live in one kit your team keeps current, and workers that use the Records app keep what they find in shared tables and lists your team can open. Your team and every connected assistant work from the same process.
• Choose what each AI worker can access. Each worker holds its own grant for each app, read-only by default, and nothing it reads can widen it. An assistant's key carries only the scopes you choose and never an app credential.
• Track what your AI workers did and what they changed. Every run leaves a receipt with its result and its cost, and every tool call is logged with its outcome, so you can see what each worker did, when, and at what price.
• Keep AI spending within the limits you set. Each worker has a per-run and a daily spending limit, under a limit for the whole account. When one is reached, new runs stop before any money moves, and the receipt says why.
• Your AI workers learn as they work and keep getting better. A language worker can check every run against its job's quality criteria. A change that makes it better waits for your approval or, in Apply proven changes mode, applies on its own once it passes a comparison on your own examples. Its runs also propose what it should remember, and you approve what it keeps.

Put a worker to work

Turn the next meeting into completed follow-through.

A worker reads your connected meeting notes, creates tasks in your task board and reports items that need your decision. Muse can use that result to plan the next step with you.

Explore the Meeting Action Item Collector, review its app requirements and connect the accounts for the job.

Hired specialists

What a worker adds beside Muse.

An AI worker is a background agent with its own instruction, memory, triggers, budget, model, app permissions and key. WorkerKit hosts the runs and Muse uses the results.

The operator pattern

Muse operates. Workers work.

Muse becomes the fleet operator. The workers do the unattended work, and Muse does the thinking about it.

For agents: search the public catalog by outcome and app, inspect a kit's requirements, then preview installation through the API. For custom classification, discover the decision source recipes and use POST /api/manage/decision-workers with your typed questions. For an exact app operation with no model, create an Action worker and call its tools (call examples).

Example: create a worker that classifies email previews into your categories. The setup guide covers deployment and funding before the first run.

What it unlocks

Background work that stays out of the chat.

Each run executes sealed inside its own context, settles its exact cost and returns a digest. The worker remembers its own state. Your conversation stays clean.

• Sealed jobs, small returns Every job runs in its own context and only the digest returns to the chat. A large crawl becomes the few paragraphs that matter, so many workers fit in one conversation.
• The right model per job Inexpensive models scan, frontier models judge. Every run is priced to the token, with reserved-versus-settled tracking, so background work is never a black box on your bill.
• A budget for each job Set per-run and daily spending limits for the worker. Review its output, status and model cost together on the run receipt.
• Keep watch on your schedule Set a cadence for monitoring and briefings in your time zone. Choose where reports arrive and which runs deserve a notification.
• Two-way work A worker can ask for a decision. Muse reads the question through the API and supplies your answer to start a linked follow-on run.
• Memory that compounds Keep useful rules and facts with the job across runs. Approve proposed memory to make it available to future work.
• Kits: install one or make your own Published kits come pre-instructed, pre-permissioned and pre-scheduled, so the job you want likely exists. When it does not, turn a tuned worker into one, or have Muse author, validate and publish one through the API.
• Create your own decision workers Ask Muse to classify email previews, score Google Sheets rows or judge calendar invitations using your own questions. It can build a decision worker through the API and bring back structured judgments for each item.
• Fleet operations Clone a tuned worker for the next client, one at a time or in bulk, with a dry-run preview. Permissions and the restriction rows behind them travel with the copy; schedules arrive disabled.
• Delivery where people live Send reports to the destinations connected to your account, including email, Slack and Teams. Muse can read the same run through the fleet API.

What workers can reach

Your apps, by explicit grant.

Connect the accounts a worker needs and choose the app operations for its job. Review that access when installing the kit and as the work changes.

Access and control

Scoped per worker, and scoped for Muse too.

Choose the app operations for each worker and the fleet operations Muse can request. Review results, change budgets and pause individual jobs in WorkerKit.

How it connects

The WorkerKit API, a key, a skill.

Muse uses the WorkerKit API at api.workerkit.ai: plain HTTPS, with OpenAPI as the protocol rather than MCP. The fleet OpenAPI document describes every operation it calls, and a scoped manager key, entered through Muse's hosted credential entry, authorizes them. It can call the API directly or use the wk CLI in its VM. The WorkerKit skill carries the operating instructions.

Follow the setup guide to verify access, choose a kit, deploy a model and check funding. Run a small job, inspect the result and cost, then configure its schedule and delivery.

Muse Code, Meta's terminal coding agent, is a different product and speaks MCP: it connects to the WorkerKit MCP server with one settings block, as described in Connecting Muse Code to your AI workers and Muse Code connectors. Both reach the same fleet.

Start with the Free plan and set a budget for model usage. See plans and runtime fees.

Questions

Muse and WorkerKit, answered.

• What is Meta Muse? Muse is Meta's personal AI agent, introduced on 8 September 2026 on iOS, Android and the web at muse.ai. It runs errands and multi-step tasks across the apps a person connects, works in a cloud computer of its own, keeps a memory, runs tasks in the background and on a schedule, and reaches outside services through connectors that Meta's Sentinel system allows, denies or sends to the person for approval.
• How does Muse connect to WorkerKit? Through the WorkerKit API, not MCP: plain HTTPS to api.workerkit.ai, with OpenAPI as the protocol. The agent reads the operations from https://api.workerkit.ai/openapi/workers.json and builds each request itself. An account admin creates a scoped manager key and enters it through Muse's hosted credential entry, and Muse sends it as a bearer credential on every call. It can also use the wk CLI in its VM with that same access.
• Can Muse read my email or CRM through WorkerKit? Muse can run a worker that uses the email or CRM access you granted to that worker, and read app-derived information in its results when the fleet scopes allow it. Configure the worker's app permissions and Muse's management and result access for the requested job.
• What does the Muse agent need to set this up? An account admin approves a scoped manager key and enters it through Muse's hosted credential entry. Muse reads workerkit.ai/muse/SKILL.md, verifies its account access and checks the selected worker's app grants, deployment, funding and budget before running. Follow workerkit.ai/muse/setup for the sequence.
• How does Muse get a worker's results? Muse reads results from the fleet feed and individual run receipts. A receipt carries the output, status and settled model cost, including structured judgments for decision workers. You can also configure reports to the delivery channels connected to your account, such as email, Slack and Teams.
• What does it cost? WorkerKit has a Free plan with 5 workers and no card required. Model usage is billed separately. Wallet-funded tokens use provider list prices; loading the wallet carries a 5.5% fee with a $1 minimum. With your own model key, the first 1M tokens each month carry no platform fee, then the fee is 5% of list-price usage. Applicable app charges are separate. See workerkit.ai/pricing for plans and runtime fees.
• Is this the same as Muse Code? This guide covers Muse, the personal agent, using the WorkerKit API. Muse Code is Meta's terminal coding agent and connects through MCP. Both can operate the same WorkerKit fleet with approved access.

Ready? The setup guide is written for the Muse agent doing the work, and the API page holds the full contract behind it.

To build a kit of your own, how to build a kit walks the procedure, and the kit-authoring skill is the file an agent keeps for it.