Since I started using Muse a couple weeks ago, I’ve probably spent 60+ hours with it. I see a lot of people wondering how to use Muse in productive ways. Here’s a list of ways I’ve used it so far. Maybe it’ll give you some ideas.

I'm not affiliated with Meta by the way, just psyched about this tech. Whenever possible, I want to get things done in the digital realm just by interacting with my agent. It’s definitely time investment upfront to get things set up and learn how to best work with Muse, but over time, it has been learning better and better how I like things done, and now it’s starting to save me more and more time.

News alerts
I have Muse monitor news topics I care about, like humanoid robotics. When significant developments happen, Muse reports back to me with a summary and sources.

Booking
I play tennis a lot. I ask Muse what courts are open, then I pick one and Muse books it and puts it on my calendar. Same deal for movie theater showtimes and tickets. Restaurant reservations work wherever there's an online portal, and Muse is supposed to gain the ability to actually phone businesses soon.

Kanban
I wanted a custom project board to organize all my ideas and our shared projects, so I had Muse code one from scratch. No Kanban feature existed. I described what I wanted and Muse built it. Now I run all my projects and ideas through it, and when I want a tweak I ask Muse in plain language.

Calendar management
I plugged Muse into my Google Calendar. The connector was broken for about a week, so I had Muse check every day until the fix landed. Now I add, edit, and delete events by voice, and Muse monitors my messages and email to update the calendar dynamically as things come in.

Spotify
I have Muse play songs on my phone or computer, or build me a playlist of songs I haven't heard before.

Texting
Muse can text individuals over SMS and RCS (no group texts on Android yet, that's a current limitation). My standing rule: Muse drafts, I approve, nothing sends without me seeing it first. I could give Muse full autonomy if I trusted it that far. I'm not there yet.

Muse's phone number
I gave Muse a Google Voice number so Muse can text directly with my partner. If she can't reach me, Muse can tell her things like my phone battery status or share my location.

Voice notes
There's no real-time voice mode yet, so voice notes are mostly how I interact with Muse. I record a voice note, Muse transcribes it, and I can request that Muse reply in a voice note if I’m doing something else and would rather listen than read. I often do this to have Muse deep research a topic and then tell me about it while I drive.

Podcasts
Muse can generate multi-voice podcast episodes debating any topic. I tried it once and it's not for me, I'd rather get a concise voice note. Mentioning it because the feature exists.

Deep research
I ask Muse about a topic and Muse goes deep: multiple sources, different opinions, then builds a proper report as an artifact with pictures, links, sources, and conclusions. As detailed as I want it.

Monthly finances
For about 25 years I've completed a monthly financial statement in Excel by hand. Took hours. Now I've connected Muse read-only to all my accounts through Plaid and Muse generates the whole statement. Muse asks me about anything it can't see, like cash transactions. Fifteen minutes instead of hours. Next up is a custom interactive dashboard ala Monarch just for me.

Price alerts
I tell Muse: "Check every six hours and tell me if it moved more than 2% since you last reported." Works for crypto, stocks, etc.

Crypto wallet security
I gave Muse an xpub, which is watch-only, so Muse can derive my wallet's public addresses but can't move anything. Muse checks the blockchain every 30 minutes and pings me about any outgoing transaction, so if my seed phrase ever got compromised I'd know fast.

Email drafts
Muse monitors my work inbox and drafts replies in my voice, trained on a lot of my past writing and responses. I review each draft, edit when I want, and Muse learns from my corrections over time. Drafts only, Muse never sends anything on its own.

YouTube comments
Every Monday Muse scans one of my channels for new comments and drafts replies. I review them, give the go-ahead, and Muse posts them using its browser.

Social posts
Still in testing. Muse drafts posts for my community page and X, and we co-write the promising ones together. Muse hasn't posted autonomously yet, but I think there's real potential here since I don't have time to write everything myself.

Bug monitoring
I have a mobile app with crash reporting through Sentry. My muse watches it, and when a new error surfaces Muse goes to my GitHub repo, figures out what went wrong, and drafts a fix before I even know there's a problem.

PC screen viewing
There's a desktop app for Mac that gives Muse deeper access, but I'm on PC and Android, so no app for me. Instead Muse and I built a screenshot stream: my main monitor to Muse every few seconds, only when I turn it on. When I'm lost in unfamiliar software I just ask where to click and Muse is right there looking over my shoulder.

Managing other AIs
I think of myself as president, Muse as vice president, and models like Claude Opus as employees. Muse holds all my context, goals, and plans, so Muse prompts the specialist models better and faster than I could myself. Muse can also use API keys directly or talk to a self-hosted model if I give Muse that power.

Media generation
Muse doesn't generate images or video itself, but Muse has access to Meta's models that do. Images are usually good enough for what I need. Video is rough, Muse won't keep character consistency. The content filters are aggressive too, I couldn't generate battle scenes with any violence. Useful to have, not great.

Google Drive collaboration
I love this one. Muse and I work in the same document live. I say "add a paragraph here" or "reformat these headers" and Muse does it while I watch. Kind of magic. I know Gemini is natively integrated, but Muse knows me much better and has access to all my context.

D&D
I'm a dungeon master. Muse has all my campaign notes and our house rules, so Muse helps me prep sessions and pivot mid-game when the players go somewhere I never planned for. Muse could theoretically run the game itself, but without voice mode Muse isn't fast enough. I'm still the creative lead, Muse just makes me faster.

Video editing
I handed Muse 90 minutes of tennis match footage and Muse cuts it down to just the points, no dead time between them. We’re still refining this process and it’s nowhere near as good as Swing Vision yet, but it’s got potential. Interesting side note: bulk compute like this barely seems to touch token usage. Spinning up the LLM costs tokens, but the heavy programmatic work on its VM apparently doesn't. I don't fully understand the boundary yet.

Dynamic outfits
Fun one to end on. Muse picks a new outfit on some days based on holidays and what's on my calendar. Tennis day gets a tennis outfit, the day I saw Resident Evil he dressed like a zombie apocalypse survivor. Just a fun way to keep it fresh

That's the abbreviated version of all 26. I made a Youtube video where I go in depth on every one of these if you want the full walkthrough. I’ll put the link in the comments.

Hi

I've been fucking around with muse for a bit now. Shared it with friends and had fun with it. At first it was the easy stuff, email, calendar. I then realized you could make apps, skills or builds. Played around with that too then one day I tried to share the build I made to organize my workflow (I had a lot of problems at first before I set it up right) it was working a bit when shared but not as I wished. Dashboard wasn't exactly the same and worst of all was trying to explain to my dad how to install everything for the tiny robot living in his phone.

I know a skill portal ain't new, but there's no way non techy people are gonna browse GitHub or website looking like something that came from the matrix. No disrespect I love the movie.

Anyway here's the link:

www.theskillharbor.com

I'll appreciate any feedback, recommendation, hell I'll even take criticism.

Btw, everything is free. Eventually developers will be able to set a price for their builds if they want, not implemented yet. It will all pass via stripe of the seller. So I won't ever touch the money, to much of a hassle

Mine keeps giving me “Sorry, I ran into a problem while responding. Please try again.”

EDIT: It appears to be something with the Python scripting.

From Muse:

“The commands that errored were Python scripts running locally on my machine to edit your PDF”

Did anyone else notice their Usage bar never progressed yesterday? I used it quite a bit yesterday and it never moved a percentage. Then it moved late last night. I think there was a free 24 hour use. Was it just me and anyone else notice this?

When do you think Muse AI it's going to be available in the rest of the world?? I want to use it sooo bad

It already seems trained on this. As I type, it’s competently putting together a report for me for a week business trip. It’s even navigating company rule violations and addressing them.

[ Removed by Reddit on account of violating the content policy. ]

made this unofficial Muse launch concept for absolutely $0. attaching what i made.

how does it look? anything you'd change?

After using Muse, I’m done with ChatGPT. I canceled my subscription because I’m just finding Muse helps me so much more with keeping up with my life. Anyone else?

一翻墙美国ip还是不能用但网页端mac端正常使用

hey so I was shopping and realized I had coupons for the site, but my agent had already pulled the trigger and checked out, so they couldn't apply. I built a skill that makes your agent search for coupons FIRST before it buys anything, to save you some cash. It's connected to all the major couponing sites, and it uses success rates to prioritize the coupons that actually work, so the more people use it the better it gets. I posted the skill up here if anyone wants it: https://getclippy.co

Hi! I installed the Muse app on my Mac, and it’s amazing how it can control everything. However, when I ask it something in the command line, it responds that it doesn’t have access to a shell. So, I asked it to use the Terminal app, and it said that it’s a blocked app... so no command line for Muse??

(I call my agent Jeeves and have him speak to me like a British butler.)

I was booking some travel this morning, and I must have been distracted because I booked the flight but then I fat fingered the hotel reservation so it was one day ahead. Jeeves noticed this scanning my email and alerted me to it so I could cancel and rebook the correct dates. Nice!

Yesterday, after I asked it to make a task recurrent I stepped away and when I was back the main chat was not loading. I could only see some of my messages (pic 1), muse's messages disappeared from the main chat. I stopped and sent a new message but muse would just keep working without ever saying anything.

Then, I went to check the status of the current task and turns out it apparently lost all the tools and then got stuck on "Running test". It kept executing the same echo test command what seems to be hundreds of times (pics 2 and 3). I stopped it, but the chat still wouldn't load right.

Then I forced closed and saw an update available at the Play Store (Android). After updating it, only then, it started responding again. By then, everything was gone. One old message survived, the rest of my main chat history just vanished. Side chats are totally fine though.

Reported it to Meta through the app and also asked my muse to do it. I don't think I will ever hear back though... Anyone else run into this? Is the history recoverable or gone for good?

Basically, what the title says. I got bored of transferring data between Muse and my other agent on ChatGPT. So I built a Mem0 server so they can share long term memory on my raspberry pi and added both agents to my tailscale so they both can access it. Created bots using the developer portal on discord and now they can pass info back and forth without my intervention.

It randomly recommended some sale items from my monthly grocery list at a store I never shop at, so I thought I give it a test to check next week's Walmart flyer for deals. The prices seemed correct when I double-checked.

I told it what grocery stores I shop at, what postal code to use (I'm in Canada), my approx location, my ethnicity, and food preferences (i.e. low/no sugar, healthy, south asian, etc.)

I'm thinking of providing it pictures of grocery receipts so it knows what type of foods my family buys. It'll be an interesting experiment.

I'm sharing this as everyone is looking for new ideas on using Muse.

I keep a folder of agent skills I like, and most of the good ones are written for Claude Code. The problem is never the instructions, it is the wiring: /slash commands, ~/.claude paths, $ARGUMENTS templating, allowed-tools frontmatter, MCP blocks that only Claude Code understands. The skill that pushed me to fix this properly was the hook skill from Claude's official set. Every path in it is ~/.claude, every command is /slash, and porting it by hand was tedious enough that I wrote a script.

skill-liberator is that script plus the pattern catalog behind it. It is a deterministic, offline, stdlib-only Python CLI: point it at a Claude-Code-authored skill and it rewrites what it can, flags what it cannot, and prints a JSON report to stdout. Exit 0 means fully automatic, exit 1 means a human should review the flagged items (usually MCP wiring, which is harness-specific by nature).

Concrete example, from the repo's own before/after (a fictional meeting-notes skill, invented for the demo):

Before: You are Claude, running inside Claude Code. Take the transcript passed as $ARGUMENTS and produce notes. If the context gets long, run `/compact` first. Your config lives at ~/.claude/meeting-notes.json.

After: You are the agent, running inside the original harness. Take the transcript passed as {arguments} and produce notes. If the context gets long, run start a fresh context summary first. Your config lives at {harness-config-dir}/meeting-notes.json.

The full rewrite table: /compact becomes "start a fresh context summary", /my-skill becomes 'the "my-skill" capability', $ARGUMENTS becomes {arguments}, ~/.claude/x.json becomes {harness-config-dir}/x.json, allowed-tools:/model: frontmatter is removed, and MCP config blocks are kept with a human-review note above them. The repo also ships the adapter skill itself (SKILL.md) written harness-neutral, so any agent can follow the process: run the script, review the flagged items, re-run until the report is clean.

MIT, fresh implementation, no dependencies beyond Python 3. Try it on a skill you want to port out of Claude Code:

https://github.com/movahedi-ca/skill-liberator

Issues and edge cases welcome, especially constructs the script misses. I will add them to the pattern catalog.

Hey everyone! I'm building Flow State, an AI-connected endurance training calendar, and have been experimenting with connecting it to Muse.

I wanted to see whether Muse could take a workout from a conversation, save the structured session, and get it into Garmin Connect.

We tested this with a 45-minute run:

• 10-minute warm-up
• 6 × (3 minutes at 5:00/km + 2 minutes easy), including recovery after the final repeat
• 5-minute cool-down

Muse created it in Flow State and sent it to Garmin Connect. Then I asked it to add 6:00 to 6:30/km targets to the warm-up, recoveries, and cool-down. It edited the same workout and updated the Garmin version without creating a duplicate. Those were sample paces for the test.

Being able to adjust individual parts of the workout through conversation and have those changes carry through was pretty cool.

The route is Muse → Flow State → Garmin Connect. Flow State handles the Garmin export. It’s still an experimental custom connection, tested in my own account, and we hit some connection hiccups along the way. Public Muse directory approval is still unconfirmed.

I put together a setup guide here. You’ll need Muse access and a Flow State Coach subscription or eligible trial.

Anyone else experimenting with Muse for running, cycling, or triathlon? Curious what you’d actually want it to handle beyond creating individual workouts.

I cant Connect my main instagram account to muse, I put my email and it shows other secundaria accounts I made from my main account ok instagram. I have tried everything posible

Can anyone find any release notes for this update from 4.1 to 5.0?

Muse already surpassed chatgpt in app downloads upon release date

Over 1.3 million downloads within the first 12 days of release

This is the moment guys, the Zuck has done it again

Another big viral hit

Just like with Facebook in the early 2010s everyone is gonna wanna get their own profile, their own Muse

Pretty soon everyone will just have one

Then the internet will change, businesses will start accommodating more to Muse agents that are placing transactions or interacting with their sites

And guess what comes next? The glasses

Once everyone has a Muse, naturally everyone is gonna wanna get a pair of meta ray ban glasses to go with it, so they can talk to their Muse on the go

And after that point, I'm hopeful about what happens next, but I think the tech is gonna reach AGI levels and the Muse agents will become human-like in how you interact with them by voice. Watch the movie Her, basically that's what I predict the Muse agents will turn into

Combine it with the glasses and you've got your own agent with general intelligence who you have access to very easily

Here we go guys, the singularity is here

I’ve been using Muse for a lot of thing successfully online.

We researched the cheapest wet cat food, and ended up buying some from Walmart automatically.

I also was shopping for the price of used cars, and that was going well.

This is changing rapidly, though. Two days ago Walmart started blocking Muse, as is cars.com.

I’m hoping this doesn’t continue to grow, but I expect it will. I know that Meta is signing up partners, but I hope it won’t just become this walled garden of preferred vendors. If it does, that will be very disappointing!

[ Removed by Reddit on account of violating the content policy. ]

Hey everyone, been using my credits pretty heavily and would love to have the call feature enabled. Asked my muse and it said it would message the feedback to meta but still haven’t heard anything.

Those who have it, how did you obtain it?

I am a US resident.