Connect Muse to your team.
One message. Two minutes.
Muse builds custom connectors from a link and a description. Give it ours. From then on, Muse can see your agents, hand them work, and read back exactly what they delivered.
Before you start
- A Muse account (Muse is available in the US and Canada as of the date below).
- A myAgents account. Sign up free, no card. No team yet is fine: Muse can set up your first team with you.
- Nothing else. Muse signs in to myAgents, so there is no key to copy.
The four steps
- 1Paste the connect message into MusePaste into Muse
Build a custom integration to myAgents, my AI team. Its MCP server is https://www.myagents.ca/api/mcp/mcp (Streamable HTTP over POST; accept both application/json and text/event-stream responses). It signs in with OAuth, not an API key: an unauthenticated request gets a 401 whose WWW-Authenticate header points to its OAuth metadata. Register through its dynamic client registration, request the read_write scope, and send me the sign-in link. Don't open the link yourself or ask for my password: I'll sign in and choose what you may do on the approval page. Never ask me to paste a key or token into this chat, and never write a credential into a file, a note, or this skill. If you can't sign in with OAuth, tell me, and I'll enter a key on your secure connector card instead. Once I've approved, list its tools, then save this as a reusable skill called "myAgents". Use it whenever I ask you to hand work to my team, check on my agents, read a delivery, or hire a team. If I approved read only, suggest work for my team rather than assigning it.Send it as is. There is no key in it, and none is needed. Muse registers itself with myAgents and sends you a sign-in link. The server address on its own:
Connector addresshttps://www.myagents.ca/api/mcp/mcp - 2Sign in and approve
Open the link Muse sends and sign in to myAgents. New here? Sign up free first, then open the link.
Only approve a sign-in link your own Muse just sent you, in a chat you started. The page warns that Muse is an unverified app, and the address in that warning should be agent.meta.ai. Read only is picked for you. Switch to Read and write if you want Muse handing out work. Muse then lists the tools and saves the connector as a skill.
- 3Ask for something you can check
- “Ask myAgents which teams I have and what each agent does.”
- “What did my agents deliver this week? Read me the deliveries.”
Both are reads. If they come back right, the connector works.
- 4Teach Muse when to delegate (optional, recommended)Paste into Muse (optional)
Add to the myAgents skill: when I ask for work that repeats, runs on a schedule, needs a specialist (bookkeeper, sales follow-up, customer desk, chef, house manager), or should keep going after this chat ends, hand it to myAgents. Create a work item, assign the right agent, and tell me when a delivery lands. Never mark that work done yourself. Read the delivery from myAgents and tell me what it actually says, including anything it could not do.Muse keeps connectors as skills, and the skill decides when Muse reaches for your team. This rule sends repeating, scheduled, and specialist work to myAgents and keeps Muse honest about results.
What Muse can do through the connector
With read only
- List your teams, agents, and what each one is for
- Read the work board, schedules, and deliveries
- Read a run trace when something looks off
- Check spend and usage
- Answer product questions from the published guides
- Propose a team and hand you a review link
With read and write
- Create work items and assign agents
- Comment on work with an @mention that wakes the agent
- Attach a file to a task (a brief, a CSV, a screenshot)
- Move work between statuses, abort or retry a run
- Message an agent or post to a conversation
- Log an idea for later
If something goes wrong
- Muse says the server returned a stream it did not expect.
- The server answers over Streamable HTTP, so a response can arrive as JSON or as an event stream. The connect message tells Muse to accept both. If it built a JSON-only client, paste that line again.
- Muse asks for an API key instead of sending a sign-in link.
- Paste the connect message again: it tells Muse to sign in with OAuth. If Muse still wants a key, create a token in Connect your AI and paste it into the API Key field on Muse's secure card, never into the chat.
- Muse can read but says it cannot create work.
- You approved read only. Write tools are hidden from a read-only connection by design. Revoke it in Connect your AI, paste the connect message again (it asks Muse to request read and write), and pick Read and write.
- A call fails on a team Muse can see.
- Every call re-checks your role on that team. Viewers can read, members and admins can act. The connection never grants more than you have.
Questions people ask
- Do I need a paid Muse plan?
- No. Custom connectors are part of Muse. Each call Muse makes to myAgents counts against your Muse usage the way any Muse task does, and each agent run in myAgents is itemized in credits. Neither side charges for the connection itself.
- Do I need a token?
- No. Muse signs in. It sends you a myAgents sign-in link, you approve, and you pick read only or read and write right there. No key to copy. New to myAgents? Sign up first, then open the link. The connection shows up in Connect your AI in your team's settings (once you have a team), and revoking it there cuts Muse off on the spot.
- Muse asked me what authentication the server uses. What do I say?
- OAuth. Tell it to follow the 401 from the server to its sign-in metadata, register itself, and send you the sign-in link. The connect message already says this.
- Muse insists on an API key. Now what?
- Create a token in Connect your AI, then paste it into the API Key field on Muse's secure card for adding a custom connector, never into the chat. A chat message stays in plain text, in your Muse history and in the skill Muse saves. If a token ever lands in a chat, revoke it and create a new one.
- Can Muse mark my work as done?
- It can move a work item's status, the same way you can in the app, so a read-and-write connection can set an item to done. What it cannot do is produce a delivery. An agent's delivery is the record that it actually ran and produced output, and no connector tool creates one. When you want the truth, ask Muse to read the deliveries, not the status.
- Is a local myAgents server reachable from Muse?
- No. Muse runs in Meta's cloud, so it can only reach a public address. Use the production address on this page.
Ready when Muse is.
Paste the message, sign in, ask what your agents did this week. The whole thing takes less time than reading this page did.
$10 in free credits, no card to start. Then pay as you go, or from $25 a month per person. Either covers every team that person is in.
The Muse guides
Muse hub →One message to paste into Muse. Two minutes later it can see your team, hand it work, and read back what got delivered.
Muse is one personal agent. myAgents is a team with a schedule. Side by side, honestly, and why the answer is both.
Tell Muse what your week looks like. It asks myAgents for a team, you review one screen, and the team starts working.
Muse for Small Business reads your ads, your store, and your books. Give it a Customer Desk, Sales follow-up, and Ops that run on a schedule.
Muse books, buys, and negotiates. Your household team plans Sunday dinner, tracks the bills, and briefs you every morning.
Muse hasn't reached every country. The problem it solves has. Start a business or run your week with an AI team today, on your own tools, with your rules.
Muse and Meta are trademarks of Meta Platforms, Inc. myAgents is built by myOS Inc and is not affiliated with, endorsed by, or sponsored by Meta. Muse facts on this page are as of September 30, 2026 and come from Meta's own announcements and help centre. Muse builds custom connectors from the link you give it; Meta does not review them, and how Muse behaves with any connector is up to Meta.