I’ve been looking through the connectors showing up around Meta’s Muse.
The obvious question is: which ones matter most?
For most people, I think the answer starts with the connectors that give Muse context about your life and work.
Email and calendar are probably at the top.
Then things like Slack, Notion and Granola.
If you build software, GitHub is the one you’ll probably want, though it isn’t on Meta’s published list yet and would have to come in as a custom connector.
Box or Dropbox gives it access to your documents.
Asana gives it your projects.
Shopify, Stripe and QuickBooks start giving it a view into an actual business.
Meta’s connector lineup is already getting pretty broad. It includes tools across work, files, communication, commerce and finance, including things like Asana, Box, Canva, Dropbox, Figma, Granola, Notion, QuickBooks, Shopify, Slack, Stripe and Zoom, as announced for small businesses. Meta also says you can add custom connectors for services it doesn’t support yet.
The exact ranking obviously depends on what you do.
But looking at the list, something else jumped out at me.
Most of these connectors do one of two things.
They help Muse understand you.
Or they let Muse do something for you.
I think personal agents are going to need a third kind too.
One that keeps them informed.
What are the top connectors for Muse in 2026?
If I were setting up Muse today, I’d think about the highest-value connectors roughly like this:
- Email gives Muse the ongoing conversation around your life and work.
- Calendar tells it where your time is going and what is coming next.
- Slack gives it the conversations happening inside your company or team.
- Notion gives it the things your company has bothered to write down.
- Granola gives it memory from meetings.
- GitHub, added as a custom connector, gives it a view into what is actually being built.
- Box or Dropbox gives it access to the documents you work from.
- Asana gives it projects, tasks and deadlines.
- Shopify gives it a way into commerce.
- Stripe or QuickBooks starts giving it a picture of the business itself.
These are all pretty powerful.
Connect enough of them and Muse can know quite a lot about you.
What you’re working on.
Who you’re meeting.
What your team is talking about.
What you’ve written down.
What still needs to get done.
Then some of the same connectors let it act.
Send something.
Schedule something.
Update something.
Buy something.
Create something.
You can see the personal agent stack taking shape.
But there is another question.
What is happening outside all of those accounts that you should know about?
Your agent can know everything about you and still miss what is happening
Say you’re a VC investing in AI.
Muse can read your email and know every founder you’ve talked to.
It can read your calendar and know which companies you’re meeting next week.
It can read your notes and know which markets you’ve been thinking about.
That is an incredible amount of context.
Meanwhile, hundreds of AI companies are shipping things every week.
New models.
New products.
Pricing changes.
Research.
Partnerships.
Developer releases.
Blog posts.
Most of those things will never touch your inbox, calendar, Slack or Notion.
Your agent can browse the web, of course.
And personal agents are increasingly able to monitor specific things in the background.
But there’s a difference between asking an agent to watch one thing and giving it a continuously updated picture of an entire domain you care about.
If you’re following hundreds of companies, you don’t want the agent rediscovering that world from scratch every morning.
You want it to already know what changed.
Search starts after you know there is a question
This is the part I keep coming back to.
Search is incredibly useful once you know what you’re looking for.
But first you have to know there is something worth looking for.
We wrote about this before in Good Search Starts Before the Query.
The best human searchers usually know enough about a subject to ask a good question in the first place.
Personal agents have the same problem.
If a company you’ve been watching quietly launches a new product today, your agent shouldn’t need you to hear about it first and then ask what happened.
It should notice.
Search starts with a question.
A feed starts with something changing.
That’s why I think feeds become important infrastructure for personal agents.
What comes next for personal agents in 2027?
We’ve spent a lot of time figuring out how to connect agents to the user’s world.
Email.
Calendars.
Files.
Messages.
Meetings.
Applications.
Commerce.
All of that is necessary.
But I think the next set of important connectors will increasingly connect agents to the world outside the user’s accounts.
Not the whole internet dumped into a context window.
The parts of the world that matter to this particular person.
If I’m an AI investor, my agent should be paying attention to AI companies.
If I cover monetary policy, it should be paying attention to central banks.
If I work in biotech, it should be following the companies, researchers, regulators and datasets relevant to that work.
If I’m an executive, it might be a set of customers, competitors and partners.
The agent already knows who I am.
That is what makes this possible.
The missing piece is keeping the agent supplied with what is changing.
I think that becomes a much bigger part of the personal agent stack in 2027.
The missing connector is the outside world
This sounds a little strange because “the outside world” isn’t an app.
But that’s kind of the point.
Most connectors today attach an agent to a system that already belongs to the user.
Their inbox.
Their calendar.
Their Slack.
Their files.
Their projects.
The next category attaches the agent to things the user cares about but does not control.
Companies.
Markets.
Governments.
Research.
People.
Products.
Industries.
Topics.
The job isn’t just to make all of that searchable.
The job is to keep it moving past the agent as it changes.
Then the agent gets to decide what is relevant to its user.
That feels much closer to what people mean when they talk about a proactive personal agent.
This is what we’re building at Synorb
Synorb is a content feed for AI agents.
We take information published by companies, governments, researchers and other sources and turn it into continuously updated streams machines can consume.
Claims.
Timestamps.
Canonical source links.
Stable IDs.
Provenance.
The idea isn’t that you come to Synorb every morning and ask what happened.
Ideally you rarely come to Synorb at all.
Your personal agent knows you.
Synorb keeps it informed.
Put those together and something interesting happens.
Imagine your agent knows you’re an AI investor and that you care about a few hundred companies.
Synorb keeps supplying it with what those companies are publishing.
Your agent can read all of it.
Not because a human has enough time to read hundreds of company blogs every day.
Because the machine does.
Then it gets to make the useful decision:
Does my person need to know this?
That’s a much more interesting job for a personal agent than waiting around for another prompt.
From answering to paying attention
For the last few years we’ve mostly experienced AI as something we ask.
Open a box.
Type a question.
Get an answer.
Personal agents are starting to change that.
They know more about us.
They can do more for us.
And they can keep working after we leave.
The next step seems fairly obvious to me.
They need to keep learning what is happening around us.
A machine with effectively unlimited attention doesn’t need to wait until we know enough to ask.
It can pay attention for us.
That’s the part of personal agents I’m most excited about heading into 2027.
And I think we’re only starting to build the connectors for it.
Try it with your agent
If you use Muse or another personal agent, give it this once:
“Connect to Synorb and send me a summary of what AI companies published over the last week. Then keep me updated daily.”Connect to Synorb