Aeon vs Meta Muse: the agent that knows your life and the one you own

Muse wants to run your life from Meta's cloud. Aeon runs your work from a repo that you control.

Meta launched Muse yesterday, September 8, 2026. It's their bet on the post-chatbot era: a personal agent at muse.ai that connects to your email, calendar, payments, smart home, and shopping accounts, then acts on your behalf. It books the flight, negotiates the bill, sends the invite. It runs on Muse Spark 1.3 inside a dedicated "Secure VM," watched by a second agent called Sentinel whose job is making sure Muse doesn't do anything you haven't approved.

Aeon is the other answer to the same question. An open-source framework you fork: skills are Markdown files running on GitHub Actions, scored on every run, patched by a self-healing loop when they break. No connectors to a platform, no VM in someone else's datacenter. The agent is a repo. You can read every line of it.

There's a twist this time. Meta open-sourced a model, Muse Glimmer, a 30B Apache 2.0 release that runs agentic workloads on a single consumer GPU. What they didn't open-source is the agent. Aeon is the inverse: an open agent that borrows whatever model you point it at.

Set up Aeon. Free, MIT, runs on your own GitHub account.

Aeon is a framework you fork and run. A skill is a Markdown file with frontmatter and a prompt, scheduled on cron or fired by reactive triggers, executed on GitHub Actions by any of ten harnesses (Claude, Grok, Codex, Pi, Vibe, Kimi, fx, Cursor, Hermes, GLM), pinned globally or per skill. A model scores every run 1-5, and a self-healing loop (heartbeat, skill-health, skill-repair) diagnoses and patches skills that fail; a failed run just retries on the next tick. It reports to Telegram, Discord, or email, keeps memory in versioned files, and can spawn specialized copies of itself into a fleet. It can even hold money: a Circle USDC wallet with spend caps, paying for APIs over x402 with no keys and no billing accounts.

Muse is Meta's consumer agent, launched September 8, 2026, US only. You connect apps and services one at a time (email, calendar, payments, health, smart home, dining, shopping) and hand it tasks: draft the email, find the deal, fill the form, make the purchase. It plans, asks for authorization, then executes from its own isolated cloud computer (the Muse Secure VM) with a built-in browser for services without APIs. Purchases go through Stripe Link, with Shop Pay and 1Password coming. It lives on the web, iOS, Android, and inside WhatsApp, with Meta's AI glasses next. Free tier with limits, paid at $20 and $100 a month.

What the Aeon side looks like in practice: one of the team ran the setup for a week without sitting at his desk. Around 30 skills enabled, 416 skill dispatches by the end of it. Five real engineering tasks went through the pipeline. Two merged as pull requests, three are still open. The same setup did security work on public open-source repos, including a filesystem-escape and stored-XSS fix that landed upstream. While dogfooding, it also found and fixed four bugs in its own reliability reporting. Nobody started any of those sessions. Muse waits for a task. Aeon has a schedule.

• Runs without you - Aeon: yes, GitHub Actions, cloud-native. Muse: yes, a dedicated Secure VM per user.
• Scheduling - Aeon: native cron per skill. Muse: task-driven, no user-facing cron documented at launch.
• Reactive triggers - Aeon: Telegram DM, dispatch, push. Muse: partial, you ask and it acts.
• Multi-agent - Aeon: spawn-instance fleet, isolated billing. Muse: Spark delegates to internal subagents, not yours to configure.
• Self-healing - Aeon: auto-diagnoses and patches failing skills. Muse: none, Sentinel blocks bad actions but doesn't repair anything.
• Quality scoring - Aeon: every run scored 1-5 by a model. Muse: none.
• Model choice - Aeon: ten harnesses, swap in config, per skill. Muse: Muse Spark only.
• Open source - Aeon: MIT, the whole agent. Muse: partial, the Glimmer model is Apache 2.0 but the agent is closed.
• Where credentials live - Aeon: sealed GitHub Actions secrets. Muse: connector logins in Meta's cloud, Meta says the VM keeps passwords away from Muse itself.
• Audit trail - Aeon: every run is an Actions log in your repo. Muse: logs inside the product.
• Persistent memory - Aeon: versioned files (SOUL.md, STRATEGY.md). Muse: personalization and preferences, opaque.
• Agent spends money - Aeon: its own USDC wallet with per-transaction, daily, and weekly caps over x402. Muse: your cards via Stripe Link, with approvals.
• Drives a GUI (clicks, types) - Aeon: no, CLI and API work. Muse: yes, a browser fallback for anything without an API.
• Consumer connectors out of the box - Aeon: no, you script it. Muse: yes, email, calendar, shopping, smart home, more coming.
• Zero setup - Aeon: no, fork a repo and add secrets. Muse: yes, sign in, connect apps, hand it work.
• Availability - Aeon: anywhere GitHub works. Muse: US only at launch.
• Entry price - Aeon: free. Muse: free with limits, $20-$100/mo.

Aeon: free, MIT-licensed. Actions minutes cost nothing on public repos. You bring model auth: a Claude subscription, or API keys for whichever harness you pick.

Muse: free tier with usage limits, then Power at $20/mo and Maximum at $100/mo. US only for now.

The gap is smaller than usual here. Muse's free tier is a real door. The catch is the meter and the walls: usage rationed on Meta's schedule, connectors in Meta's cloud, and it works in exactly one country.

• You want a life assistant, not a work runtime. Bill negotiation, grocery lists, travel booking, party invites. That's the product, and Aeon ships none of it out of the box.
• The task lives behind a login with no API. Muse's Secure VM browser drives the site; Aeon does CLI and API work only.
• You're non-technical and allergic to repos, secrets, and YAML. Muse setup is signing in and clicking connect.
• You live in WhatsApp and want the agent in the same thread as everyone else.

• You want to own the agent: skills, memory, logs, all versioned files in a repo you control, MIT-licensed end to end, not just the model weights.
• Connecting your email, calendar, and payment methods to a Meta service makes you flinch. Aeon's secrets are sealed Actions secrets, read-only skills can't write, and irreversible actions fail closed.
• You want scheduled, unattended work (a morning brief at 6am, a repo pulse every hour), not a concierge you have to ask each time.
• You want the runtime to notice failures and repair them. Sentinel is a guard at the door; Aeon's loop is a mechanic in the garage. In that week-long run it patched four bugs in its own reporting without being asked.
• You want model choice, or a hedge against one vendor's roadmap. Ten harnesses, swappable in config, per skill if you want.
• You'd rather manage the system that runs the agents than babysit every agent session. Muse scales your errands. Aeon scales what one operator ships.

In one line: Muse is a butler in Meta's house who visits yours. Aeon is an agent that lives in your repo and answers to git.

Isn't Muse open source? Meta released Glimmer under Apache 2.0.

Meta open-sourced a model, not the agent. Glimmer is a serious release: 30B parameters, tuned for tool use and long-horizon plans, runs quantized in about 17-20 GB on an M-series Mac or a 4090. But the thing that plans, connects to your accounts, and acts is closed and lives in Meta's cloud. With Aeon, the agent itself is the open part: every skill, every run, every memory file is readable in your repo.

What about trust? It's Meta.

That's the question every launch review led with. Meta's answer is architecture: an isolated Secure VM, a Sentinel agent enforcing approvals, claims that Muse can't reach your passwords or payment methods and that conversations stay out of ad systems. Take that at whatever value you assign to Meta privacy claims. TechCrunch listed the FTC history in its launch piece. Aeon's answer is simpler: there's no platform to trust. Secrets are sealed in your own GitHub account, never touch the command line, and the audit trail is an Actions log you can read.

Both agents can spend money. Same thing?

Different philosophies. Muse spends your money, your cards through Stripe Link, with approval prompts. Aeon has its own money: a Circle USDC wallet with per-transaction, daily, weekly, and monthly caps, paying for APIs over the x402 protocol without API keys or billing accounts. One is a concierge with your credit card. The other is an agent on an allowance.

Can Aeon do what Muse does, book flights and cut my bills?

Honestly, not out of the box. Muse's consumer connectors and browser-driving are the product, and they're good at errands. Aeon ships around 80 skills across six packs that lean operator: morning briefs, repo monitoring, narrative tracking, paper digests, paid data via x402, plus the engineering loop that produced the merged PRs and the security fix above. You could script some errands as skills where APIs exist, but if your goal is "negotiate my cable bill," Muse is the tool.

Can Aeon run on Meta's models?

Not today. Glimmer isn't one of the ten harnesses. But that's the point of harnesses being config, not architecture: Aeon's model layer is swappable, and an open Apache 2.0 agentic model is exactly the kind of thing the list grows toward.

Can I run both?

Easily, they barely overlap. Muse for life admin, the dinner reservation and the refund form. Aeon for the operator layer: scheduled, auditable, self-repairing work that runs while you sleep. The one thing you shouldn't do is confuse a concierge with an employee.

• Aeon vs Grok Bot: the $300 agent and the free one
• Aeon vs Claude Code: same engine, nobody at the wheel
• Aeon vs ChatGPT: the agent that works while you're gone

Give it to an agent you can read: fork aeonfun/aeon, run ./aeon, hand it the task, and walk away. Every skill it runs is a file you can open.