Is Meta Muse free? What it changes for job hunting

Is Meta Muse free now, and what did Meta actually open source?

Muse already had a free tier — it has since launch on 8 September 2026 — so the honest answer is that nothing about the price changed on 2 October. What Meta gave away that day was the hardware code: open-source ESP32 firmware and a Linux SDK, released under Apache 2.0, so anyone can build a physical device that talks to Muse. Nat Friedman, head of product at Meta Superintelligence Labs, announced it as Muse Gadgets. The headlines compressed that into "Meta is giving Muse away for free," which is close enough for a social post and misleading if you're making decisions off it. The agent itself still runs on Meta's servers, and your gadget needs a token to reach it.

The release is deliberately small and hacker-flavoured. The repository splits into directories for ESP32 boards, Linux devices and skills, each with a readme plus a file aimed at coding agents. Meta published blueprints to copy: a gadget with a 1.75-inch AMOLED display, an HDMI stick that takes over your TV, a remote-control form factor, a mini e-ink tablet. The showcase builds run on a Raspberry Pi and off-the-shelf dev boards. There's a Discord channel for support, and a free USB-C accessory called Muse Home Link that bridges Muse to your home network — 5,000 units, free to US subscribers while supplies last. Meta's own framing: built by hackers, for hackers, with bricked boards among the likely side effects.

The distinction worth holding onto is free-as-in-build-it versus free-as-in-weights. Nobody can download the Muse model and run it offline; every gadget pairs through an API token, and every request still travels to Meta's service. Access to the app sits on three usage tiers — a free one, Power at $20 a month and Maximum at $100 — with the same features on each and different volume allowances rather than feature gates. It's US-only and 18-plus for now, and reports say a payment card is required even on the free plan. So "open source Meta Muse" overstates what happened. What genuinely got cheaper is wiring a capable agent into something that isn't a phone.

Does an open-source Muse change how your CV gets screened?

Not yet, and not directly — no applicant tracking system is going to run on hobbyist ESP32 firmware. What changes is the cost of putting an agent somewhere it wasn't before, and hiring is one of those places. The screening you're already up against doesn't come from gadget SDKs; it comes from ATS platforms with AI filtering switched on, and from recruiters pasting your CV into a chat window between meetings. Read the gadget release as a signal rather than a mechanism: agent plumbing is turning into commodity infrastructure. The practical upshot is that the tools reading your CV will get more numerous and more varied over the next year, not fewer and more standardised.

The adoption numbers already point that way. Industry surveys put AI use at roughly half of organisations with any recruiting technology, with CV screening consistently among the top two or three applications, usually just behind writing the job ad itself. Near-universal ATS use at the largest employers has been true for years. The gap was always smaller companies: mid-sized firms adopt some form of automated screening at roughly double the rate of businesses under a couple of hundred people, because integration cost money and attention that nobody had spare. Meta pushed Muse into small business on 29 September 2026 at the same pricing as the consumer app. Draw the line between those two data points yourself.

For you, this has one practical consequence: you can't design for a single known reader any more. Your CV might be parsed by an enterprise ATS, summarised by a general-purpose assistant, or skimmed by a hiring manager who asked an agent for the three strongest applicants and never opened your file at all. Those readers fail differently, and they all fail silently — no bounce-back, no explanation, just nothing. My blunt opinion: the most expensive mistake in job hunting right now isn't weak wording, it's a beautiful two-column template with icons and text boxes that collapses into word salad on extraction. Build for the dumbest parser in the chain and every smarter reader still gets the full picture.

Can you run your job search on a Muse assistant, and should you?

Yes for the mechanical half of a job search, and no for the half that decides whether you get hired. Muse was built to book travel, fill in forms and buy things on your behalf, which maps neatly onto application admin: tracking which postings you've already seen, filling repetitive fields, keeping deadlines straight, drafting a first version of anything. Where it struggles is judgement — knowing which slice of your six years actually answers this particular ad, and what a specific hiring manager will quietly discount. That split is worth being strict about, because an agent sounds equally confident in both halves. Treat it as the assistant who handles logistics, not the coach who decides your positioning.

A few habits make these assistants genuinely useful. Paste the job ad as text rather than a link — LinkedIn and Indeed block automated fetching, so a URL usually earns you a confident summary of nothing. Ask for the requirements as a plain checklist, then mark honestly which ones you can evidence with a number, a named system or a shipped outcome. Use it to generate likely interview questions from the ad, to compress a nine-line responsibility blob into two results-led bullets, and to keep a simple tracker with dates and contacts. Hold one master CV as your source of truth and tailor from it, rather than letting the agent regenerate everything from scratch each time and drift away from the facts.

The failure mode is predictable. An agent optimises for text that reads well in a chat window, which isn't the same as text that survives extraction, and it can't see what your exported PDF actually looks like — whether your dates landed in a sidebar the parser skipped, or your skills ended up inside a graphic. It also writes with a recognisable rhythm: stacked triads, "spearheaded", vague scale words where numbers belong. Recruiters have started clocking it. If you've been drafting with an assistant, check whether your CV reads as AI-written before you send it anywhere that matters. Mass auto-apply, meanwhile, is a volume strategy with a response rate to match.

Is Meta Muse safe for my CV and personal details?

Assume anything you hand a consumer AI agent becomes stored data with a memory attached. That isn't a Muse-specific warning, but Muse makes it concrete: it keeps memory across sessions, connects to services including your inbox and messaging, requires a payment card on file even for the free tier, and is limited to US users aged 18 and over. A CV is a dense packet of personal information — home address, phone number, full employment history, occasionally a detail in the footer you forgot was there. Uploading it to an agent isn't reckless, and plenty of people do it daily. It's just a decision, and most make it without ever noticing they made one.

Sensible precautions cost you almost nothing. Cut your street address down to city and country, keep one email address you're happy to see circulated, and strip referees' phone numbers until an employer actually asks for them. Check what the assistant retains in memory and clear it when a search ends. Don't connect a work inbox to an agent that's helping you leave that job. If you grant an assistant permission to act — book, buy, submit — read the approval prompt before you wave it through rather than after, because an application filed to the wrong company isn't something you can quietly undo. And keep the DIY gadget angle in proportion: that SDK ships as-is under Apache 2.0, with no warranty.

The risk I'd rank first isn't a leak, though — it's outsourced judgement. An agent will cheerfully tell you your CV looks strong because the prose is clean, while having no view of the things that actually get you filtered: how the file extracts, whether your job titles map to the vocabulary this employer uses, which skills named in the ad you never mentioned once. That's a measurement problem, and a chat window is the wrong instrument for it. A free CV analysis that scores structure and ATS compatibility answers it in about a minute, and it reads the file you'd actually send rather than a pasted transcript of it.

How do you check your CV reads well to any AI screener before you apply?

Run the exact file you plan to send — same format, same template, same export — through a check that scores how it extracts, not just how it reads. HRLens's free CV analysis accepts a PDF, a Word document or an image, with Hebrew OCR supported, and returns an overall score out of 100 rated Weak, Decent, Strong or Excellent. You also get five category scores covering experience, tech stack, impact and ownership, clarity and structure, and ATS compatibility, plus a list of your genuine strengths and a visual layout analysis showing what your design is doing to your content. That last piece is precisely what a chat assistant structurally cannot give you.

Then make it specific to the job in front of you. Paste the job description into the analysis — paste the text rather than a URL for LinkedIn and Indeed listings, since both block fetching — and you'll see which skills the ad asks for that your CV never mentions, with suggestions aimed at that role instead of generic advice. Every analysis also generates ATS-friendly rewritten versions of your CV in six templates that mirror the ad's own language. If you want the harder feedback, the paid Full Analysis at $3.99 adds the fix list, the reasons you might be rejected, your blind spots, an estimate of interviews per ten applications against the market, and a salary estimate.

A workflow that holds up in 2026 looks roughly like this. Let an agent handle breadth — finding postings, summarising ads, chasing follow-ups, keeping the tracker tidy — and run a dedicated CV check on the artefact itself every time the template or the target role changes, because those are the two moments things break. If the file is past patching, rebuild instead of editing: the chat-based CV builder takes an old CV or just a description of your experience, refines it by conversation, and exports to PDF or Word across six templates, writing Hebrew CVs in Hebrew with right-to-left export. More screeners than ever will read your CV this year. Know what they see first.

Frequently asked questions