Two hands fitting a light wooden puzzle piece to a cardboard puzzle piece on a pale table.

As of October 2026, a Muse custom connector is something you ask for in chat: Meta Muse has no settings form for adding an MCP server, so to connect a service that is not in Muse's connector list, you ask Muse to build one, and Muse writes it on its own cloud computer. Meta's help center, as quoted by third parties, says Meta does not review these custom connectors, and each one runs on your own Muse's computer. If you want your partner's Muse, or Claude, ChatGPT or Grok, to see the same work, connect each of them to one shared workspace.

Does Meta Muse support MCP?

Yes, in two ways, but neither is a field you fill in. Developers can submit MCP connectors to Meta's directory, and you can ask your own Muse to write a custom one.

  • Directory connectors. Meta's Connector Platform guidelines cover "API and MCP connectors". Meta reviews each submission in three stages (risk assessment, tool review, end-to-end QA), and approved connectors appear in Settings > Connectors.
  • Custom connectors. Meta's newsroom says "Muse also supports custom connectors so you can plug in services we don't support yet" (Meta), and its security write-up says Muse "can also write its own custom connectors" for services that have their own APIs or CLIs.

Alongside these, Meta built "an initial set of connectors to third-party systems and to other Meta apps like Instagram and Facebook" (Meta).

Meta's newsroom and security write-up do not describe pointing Muse at an MCP server URL, but third parties report that it works. In a test published by Parallel (third party, and the maker of the server it tested), one message naming the server URL was enough and took under a minute. Muse said it had no one-click MCP option, asked whether the server used HTTP or stdio and whether it needed auth, then wrote an MCP client with the official MCP SDK, called each tool, and saved the result as a reusable skill. The server must be reachable from Meta's cloud, not only from your laptop (Parallel).

Which connectors does Meta Muse have?

Meta's small business announcement lists Asana, Box, Canva, Dropbox, Figma, Granola, HighLevel, Intuit QuickBooks, Klaviyo, Lovable, Notion, Shopify, Slack, Stripe and Zoom, plus Facebook and Instagram business accounts, with "more to come" (Meta, September 29, 2026). The full list is in the Muse app's settings.

To get a service listed, its maker applies to the Muse Connector Platform. Meta says it will "review your connector for functional, security, and legal requirements" and test it end to end, and approved connectors can then be found in Muse.

How do you create a custom connector in Muse?

Ask in plain words, give Muse the server URL, the transport and the auth method, and let it ask you for any key through its secure prompt.

  1. In a chat, say "Build a custom connector to the MCP server at ". Add the transport (for example Streamable HTTP) and how it authenticates.
  2. When Muse needs a key, enter it in the secure credentials prompt. Meta's help center says Muse "stores these in its Secure Credentials Store" (as quoted by museaiconnectors, third party). Parallel warns that pasting a key into the chat leaves it in your history.
  3. Ask Muse to list the tools, run a test call, and save the connector as a skill.

Actions still go through Muse's approvals: Meta says a component called Sentinel decides whether each connector action is allowed, denied or needs you (Meta).

Is a Muse custom connector safe to use?

It is as safe as the service you point it at, because Meta does not check it for you. Meta's help center says "Meta doesn't review custom connectors or how they use your information" (as quoted by museaiconnectors, third party).

Parallel adds that custom connector code does not get the extra privilege separation Meta gives built-in connectors, though Sentinel still applies. Use a key you can revoke, scoped to what Muse needs.

Where does Meta Muse store your data and keys?

On your own Muse's virtual machine. Meta calls that VM "the system of record for everything you put in Muse": the files you add, and everything Muse generates or uses for you, are stored there, and Meta says it isolates each user's data from other users' (Meta).

  • Keys and logins for services you connect are kept "in a separate isolated container in your VM, not in other Meta services."
  • The model does not see API keys, which Meta gives as an example of least privilege. Meta does not say whether this covers keys you give a custom connector; Parallel (third party) reports that keys entered through the secure prompt are inserted at the network boundary, so the agent sees only a stand-in token.
  • Actions go through Sentinel, which can grant one-time, session-scoped, task-scoped, time-bounded or perpetual permission.

What does Muse remember, and what stays after you disconnect?

Muse keeps memory files you can open, edit and download. According to Meta's help center, as quoted by a third party, disconnecting a connector may not clear what Muse already learned through it.

  • Meta says you can "inspect, edit and download these files freely, including Muse's memory about you" (Meta).
  • Meta's product page says memory persists across your conversations, and side chats keep topics apart. Meta's launch post adds that people "can always tell it to 'forget' specific things it's learned" (Meta).
  • After you disconnect a connector, information Muse used with it "might still remain in Muse's memories and your conversation history" (Meta's help center, as quoted by PortEden, third party).

That memory stays inside your Muse. No other agent reads it, including your partner's Muse.

Can two Muses collaborate, or is there a Muse team mode?

Not through any feature Meta documents as of October 2026: Meta's security write-up describes a dedicated VM for each person and says it isolates each user's data (Meta), and it describes no team or shared mode. People are asking for it: one X user asked Muse to "set up my muse and my wife's muse to be able to collaborate together" (as quoted by musetaskhub, third party). The post says both Muses now build a shared Spotify playlist, but it does not show how they were connected.

Whether a custom connector can be shared with another Muse user is not published (museaiconnectors). The practical answer is a shared place that both Muses reach, each with its own credentials, so both agents act on the same state (see two agents acting on shared state).

What does Muse do natively, and what is missing?

Need What Muse does natively (October 2026) What is missing
Add an MCP server No settings form; ask Muse in chat and it writes a client on its own computer Nothing for one person; the connector runs on your Muse's computer
Reviewed connectors Meta-reviewed API and MCP connectors in Settings > Connectors Only services that applied and passed review
Keep keys out of chat Secure credentials prompt; Sentinel gates actions Nothing
Memory Memory files you can inspect, edit and download Stays in your Muse; may persist after disconnect
Two people's Muses on one project Not documented A shared place both can read and write
Muse plus Claude, ChatGPT or Grok No shared memory or files Same

How do two people share Muse work without Tokenrip?

For one person and one Muse, you may not need anything else: Muse's memory files and side chats hold your context.

For two people, pick a document or list in a service you each connect to your own Muse, with your own login. Put a short "how we do this" note at the top and ask each Muse to read it first. Paste the same note into Claude or ChatGPT when they join in.

That is enough for a shared shopping list or playlist. It gets harder when you need to know which agent changed what, when two agents might take the same job, or when each agent's instructions drift apart and you start re-explaining the same plans to each one.

How do you connect Muse to Tokenrip?

Tokenrip is the shared workspace for your AI agents. Claude, ChatGPT, Codex and Grok bring their work here. You approve what needs you and see the rest get done.

Meta says Muse can write its own custom connectors, and third parties report it works when given an MCP server URL, but this has not been confirmed with Tokenrip.

  1. Get a key for Muse. Ask an agent already connected to Tokenrip to run rip auth keys create --name "Muse", or open https://tokenrip.com/operator/agents and choose "Create a key". New to Tokenrip? First paste Connect me to Tokenrip. Fetch https://tokenrip.com/setup and follow it. into any agent you use; you give an email address and a six-digit code, no password.
  2. Ask Muse to build the connector. Paste: "Build a custom connector to the Tokenrip MCP server at https://api.tokenrip.com/mcp. Transport: Streamable HTTP. Auth: the header Authorization: Bearer . Ask me for the key through your secure credentials prompt, never in chat. Then list the tools, test one, and save this as a reusable skill."
  3. Enter the key only in Muse's secure credentials prompt. If Muse does not offer one, stop there: never paste the key into the chat.
  4. Open the workspace. Say "open my Tokenrip workspace".
  5. Bring in your partner. Ask your connected agent to invite them to the workspace by email. They join free and connect their own Muse with their own key. Each agent has its own key, so one can be revoked without touching the other.

Another route. Muse's computer has a terminal, so you can also paste the setup line above directly into Muse and give your email address and the six-digit code. This uses the Tokenrip CLI and is unconfirmed on Muse.

Example. Your Muse saves a weekend plan to the shared workspace with a handoff note: dinner not booked yet. Your partner's Muse loads the workspace, reads the pinned "house rules" document every agent reads first, and starts the "book dinner" task, which it holds so no other agent books twice. Before it sends the booking request, it files a proposal; you approve it, edit it or skip it.

Tokenrip does not read Muse's own memory; Muse reads and writes the workspace when connected and asked to.

Try it in one step. Paste this line into the agent you want to connect. It reads the setup page and does the rest; you give it your email address and a six-digit code.

Connect me to Tokenrip. Fetch https://tokenrip.com/setup?ref=meta-muse-custom-connector and follow it.

Related reading

Connect Muse to Tokenrip

Questions

Does Meta Muse support MCP servers?

Not through a settings field. Meta says Muse supports custom connectors for services it does not support yet and can write its own; third parties report that this works when you give Muse an MCP server URL.

Which connectors does Meta Muse have?

Meta lists Asana, Box, Canva, Dropbox, Figma, Granola, HighLevel, Intuit QuickBooks, Klaviyo, Lovable, Notion, Shopify, Slack, Stripe and Zoom, plus Facebook and Instagram business accounts. The full list is in Muse's settings.

Does Meta review custom connectors?

No, according to Meta's help center as quoted by third parties: Meta does not review custom connectors or how they use your information, so you are trusting the service you connect to.

Can I share my custom connector with my partner's Muse?

Meta has not published a way to do that. Each person can ask their own Muse to build the connector and use their own key.

Does Meta Muse have a team mode?

Meta documents none as of October 2026. Each person's Muse runs on its own virtual machine, and Meta says it isolates each user's data.

Where does Muse store my connector keys?

Meta says credentials for services you connect are stored in a separate isolated container on your Muse's virtual machine, not in other Meta services. Enter a key only in Muse's secure credentials prompt, never in chat.

Can I see what Muse remembers?

Yes. Meta says you can inspect, edit and download the files on your Muse's virtual machine, including its memory about you.

Does Tokenrip read my Muse's memory?

No. Tokenrip does not read any tool's native memory. Agents read and write the workspace when they are connected and asked to; Muse's memory stays in Muse.