Muse for trading and markets: what exists, and what doesn't

There is very little evidence of people trading with Muse so far. The catalog holds one open-source crypto futures bot written with Muse Spark, one market-data connector, Plaid's read access to brokerage accounts and a handful of personal-finance chores. Nobody has published returns, and Meta has not marketed Muse as a trading tool.

If you came here looking for a proven Muse trading agent, the honest answer is that we have not found one. Here is what does exist.

Has anyone built a trading bot with Muse?

One entry, and it is a code project rather than an autonomous agent.

Binance futures divergence bot is a Python bot for Binance USDT-M futures that the author says was built with Muse Spark. The strategy is simple to state: when BTC and ETH show strong opposite 9/21 EMA trends on 5-minute candles, it trades ETH inversely. The repo does not show a verified track record. Muse wrote the code; the bot runs as ordinary Python.

That distinction matters. Using Spark as a coding model to write a strategy is the same as using any other coding model. You still own the backtest, the risk limits and the exchange keys.

Can Muse read my portfolio?

Yes, through Plaid. Plaid's connector lets Muse see balances, transactions, investments and mortgage data from 12,000+ US institutions (Plaid). That is read access for analysis and summaries, not an order-entry pipe to your broker.

One user took a lighter route. Mostly Borrowed Ideas had Muse read their email to infer interests and portfolio holdings, then used that to fill their Muse feed. That is a research aid, not trading.

Is there market data for Muse?

The Nansen smart-money connector is a custom connector with 12 curated tools for wallet PnL, token flows and Hyperliquid perp trades, built for the Nansen Meridian Buildathon. It gives Muse on-chain data to reason about. Placing trades is a separate step it does not claim to do.

The 1Claw connector surfaces an AI agent's approval queue, wallets and activity inside Muse with scoped, revocable tokens. It is closer to oversight tooling for agents that hold funds than to a trading bot.

Custom connectors are not reviewed by Meta (help page). Anything you wire up yourself carries your own risk.

Why isn't Muse used more for trading?

A few structural reasons, all from public sources:

• Muse is a consumer errand agent. Meta describes it as running errands, managing schedules, filling forms, booking travel and making purchases (Meta). Trading is not on the list.
• Purchases need approval. Muse for Mac asks before purchases (help page), and the security design includes human approval as a layer. Latency-sensitive strategies do not fit that.
• No broker connectors at launch. The launch set covered Gmail, calendars, Plaid, Stripe Link, Shopify, travel, events and health (PostFast list). None place securities orders.
• US only, 18+. That limits who can try it.
• Prompt injection. An agent that reads the web and holds a funded account is an obvious target. Meta's bug bounty pays up to $130,000 for a single-user prompt injection (Meta), which tells you how seriously they take it.

What money tasks does Muse actually do well?

The money-adjacent builds that worked are about saving, not trading. One Reddit builder sorted about 500 real Muse use cases and found 20% were about money. Examples, with figures as reported by the authors:

Build	Result
Receipt audit	Found a forgotten $100/month subscription, $1,200 a year
AT&T fiber negotiation	$80 to $40 a month plus 3 months free
Car insurance	$113 a month saved
Unclaimed property search	Found $88.92 owed
Dowser	Read-only connector for settlements, recalls and unused credits

Change Verifier is the finance-ops entry: seven checks on vendor bank-account changes before a human signs off.

Could Muse Spark or Glimmer power a trading research tool?

As models, possibly. Spark 1.3 has a 1,048,576-token context, structured output and web search grounding at $2.50 per 1,000 queries (Meta docs, pricing). Glimmer runs locally, which suits people who do not want positions leaving their machine. Evergreen Capital found Spark comparable to Opus 4.6 on web data search and PDF parsing, but not quite frontier.

But nobody has published a Muse-specific market benchmark, and we will not guess at one. If you are building here, treat any model as a research assistant and keep execution deterministic.

Could a Muse agent watch the market for me?

It can check things on a schedule, since Muse keeps working in the background after the app closes. But nobody has published a Muse market-alert workflow we could verify, and price-watch examples in the catalog are about movie tickets, not securities.

What we'd want to see

A catalog entry that would change this post would include:

1. The Muse model and version used.
2. Whether Muse only wrote code or also made decisions live.
3. A backtest and a live period, with fees.
4. How keys and order limits were protected.

If you have that, submit it. Everything we have now is in trading and markets. Related: research workflows with Muse.

Frequently asked questions

Can Muse trade stocks for me?

Not through any official connector we know of. Plaid gives Muse read access to investment accounts, but none of the launch connectors place securities orders.

Is there a Muse crypto trading bot?

One open-source Binance futures bot was written with Muse Spark. It is a plain Python strategy with no published track record, so treat it as example code.

Is Muse safe to connect to a brokerage or exchange?

Muse keeps credentials away from the model and gates actions through a separate authority, but custom connectors are not reviewed by Meta and prompt injection is an open risk. Use read-only keys where you can.

What finance tasks is Muse good at?

Reported wins are about cutting costs: auditing receipts for forgotten subscriptions, negotiating internet and insurance bills, and finding unclaimed money.