How to use Meta Muse to clean up your social media

By Dave Fradette · Published October 23, 2026 · Last reviewed October 23, 2026

Short answer: Muse, Meta’s personal AI agent, can work in a web browser and act for you once you approve it, so it can help request your data exports and work through old posts and likes. Give it a specific list instead of “clean up my account,” set permissions to ask before each change, and sign in to each platform yourself. PruneSocial’s report includes a cleanup file built for exactly this. PruneSocial isn’t affiliated with Meta.

What Muse can do today

Meta launched Muse in the US on September 8, 2026, on iOS, Android and muse.ai. Meta says it “doesn’t just answer questions, it actually does the work,” and that it can open a browser and fill out forms.^[1] It runs on a separate computer with its own browser that Meta calls Muse Secure VM, and Meta says Muse has no visibility into your passwords.^[1] If a service has no connection Muse can use, it can reach the service through that browser.^[2]

For Facebook, Instagram and Threads, the picture is narrower than the marketing. Meta’s built-in connection is documented for reading content and account context, and as of mid-October no Meta documentation described Muse deleting your posts.^[3] One tester got Muse to post in a Facebook Group only after letting it open Facebook in its own browser and signing in there himself.^[4] Nobody has published a test of Muse deleting old posts or removing likes yet, so treat that as something to try carefully, not a feature.

Set it up safely first

Make it ask every time. Muse’s permission settings default to “Ask for some actions” or “Always ask,” and when it asks you can choose Allow once, Allow for this task, Allow for this site, Always allow, or Deny.^[3] For a cleanup, use Allow once or Allow for this task. Don’t pick Always allow.
Know why that matters. In October a creator who let Muse run a Marketplace sale traced his home address being shared with a buyer to “always allow” permissions he had granted when he connected the account.^[5]
Sign in yourself. When Muse opens a platform in its browser, take control and sign in on the platform’s own page. Never paste a password into the chat.^[4]
Check its log. Muse keeps an activity log of actions and the permissions you granted.^[3]

Step by step: Muse plus PruneSocial

Request your exports. Copy the “Request my exports” prompt from the agent page into Muse and list your platforms. The export steps for each platform are written so an agent can follow them. Exports take minutes to days to arrive.
Decide where the files go. If Muse downloads your exports, they land on Muse’s own computer, not yours. If you’d rather they didn’t, download them yourself from the platform’s email.
Run the scan. Open prunesocial.com/scan, tick the two consent boxes yourself, and add the ZIPs. The free preview scores 50 items. The $49 report scores up to 5,000 from the last seven years.
Hand Muse the cleanup file. Your report’s download page includes PruneSocial-Agent-Cleanup.json and a ready-made prompt. Upload both to Muse. Each item has the platform, date, why it was flagged, the suggested fix, and a direct link when your export included one.
Approve one item at a time. Muse should show you each item and wait. Keep, edit, delete or skip. At the end, ask for its log.

Platform notes

X: X’s rules say you may not like posts in an automated manner, and warn that scripting the X website “may result in the permanent suspension of your account.”^[6] The safer way: let Muse open each item, and you click Delete or Unlike yourself.
Facebook: Facebook’s own Manage Activity tool in the Activity Log can archive or trash posts in batches.^[7] Facebook’s export doesn’t include links to each item, so the cleanup file gives Muse a “find it” hint (Activity log, type and date) instead.
Everything else: Go slowly and stop at any warning or CAPTCHA. Muse can make mistakes. Meta itself warns it “may be inaccurate or take unexpected actions,”^[3] which is why every change should be one you approved.

Why a list beats “clean up my account”

An agent scrolling your profile sees what’s on screen, not years of likes and comments, and it has no screening rubric. Your exports hold the full history. PruneSocial reads them, scores the text against categories screening vendors publish, and turns the results into a short list. Deleting everything isn’t the goal either. A normal, human profile usually reads better than an empty one.

Common questions

Does PruneSocial connect to Muse? No. PruneSocial never signs in to your accounts or connects to Muse. You give Muse the files from your report, the same way you’d give them to a person helping you.

Can Muse tick the consent boxes on the scan page? No. Those confirm you’re 18 or older, the accounts are yours, and you consent to health information in your text being processed. Tick them yourself.

Is Muse available where I live? Meta’s announcement lists a US rollout.^[1]

Give your agent the right list

PruneSocial scores the text in your own exports against published screening categories. Every paid report with findings includes a cleanup file your agent can work from, one item at a time, with your approval. Try a free 50-item preview first.

Start a free preview