Walnut vs M1 Finance: AI Baskets or Pies?
Walnut and M1 Finance are the two best-known apps for building basket-style portfolios in the US. They tackle the same problem in opposite directions: M1 is the broker, with manual UI-driven pies and scheduled auto-rebalance; Walnut is an AI layer on top of your existing broker, with baskets built in conversation. Here is the honest comparison, including where M1 genuinely wins.
At a glance
| Walnut | M1 Finance | |
|---|---|---|
| Is it a broker? | No, sits on top of your broker (via SnapTrade) | Yes, M1 is the broker |
| Portfolio model | Thematic baskets, AI-built in conversation | Pies, manual allocations in the UI |
| AI assistant | Built-in agentic AI + Claude / ChatGPT support via MCP | No conversational AI |
| Auto-rebalance | Manual trigger via the Invest dialog | Scheduled (set-and-forget) on M1 Plus |
| Web search in chat | Yes, for current market data | No |
| Custodial accounts (UGMA/UTMA) | Inherits broker (varies) | Yes |
| Cash management | Inherits broker | M1 Spend (debit card + checking) |
| Pricing | Free | Free; M1 Plus $10/mo (margin + cash management perks) |
| Works with your existing broker | Yes, any broker SnapTrade supports | No, must use M1's brokerage |
| Margin | Inherits broker | Yes, M1 Borrow / Margin Plus |
The architectural difference (and why it matters)
M1 is a broker. Your money lives there, your trades clear there, your tax forms come from there. Walnut is not a broker; it connects to your existing one through SnapTrade and adds intelligence on top. That single difference cascades into most of the comparisons below.
If you want your basket app and your broker to be the same app, M1. If you want to keep the broker you already use (and not have to ACATS-transfer assets) while still getting basket-style management plus AI, Walnut.
Building a portfolio: Pies vs AI-built baskets
M1's Pies are the more mature product: a polished visual builder where you drag-and-drop allocations across slices, clone Expert Pies, or layer Pies inside Pies. It's the gold standard if you already know what you want.
Walnut's baskets are built in conversation with the AI. You describe a thesis (“AI infrastructure”, “dividend growth large-caps”, “off-price retail in a tightening consumer”), the AI researches the relevant constituents, proposes target weights, and you approve. It's a different shape, better for exploration, worse for “I already know exactly what allocations I want”.
Neither is strictly better. They map to different mindsets: top-down recipe vs. conversational research.
Rebalancing, scheduled vs. AI-assisted on demand
M1 has scheduled auto-rebalance, set a cadence (weekly, monthly, quarterly), forget about it, drift gets corrected automatically. This is genuinely one of the best things about M1.
Walnut shows you the exact trades that would bring your basket back to target weights and lets you review and execute through the Invest dialog. The AI can also explain why drift happened (one position rallied, another sold off) and walk through whether the imbalance reflects the thesis or just noise. More deliberate; not set-and-forget.
If you want hands-off, M1. If you want AI-assisted thinking before each rebalance, Walnut.
AI, Walnut's clearest advantage
M1 doesn't have a conversational AI. Walnut has a built-in agentic assistant (Claude Sonnet under the hood) that can call tools to read your real portfolio, build or edit baskets, run drift and performance analysis, search the web for current market data, and answer follow-up questions across the whole conversation. It also exposes an MCP connector so Claude Desktop and ChatGPT can talk to your portfolio directly if those are your preferred AI tools.
If AI is incidental to how you invest, this doesn't matter. If you want AI as the primary interface to your portfolio, Walnut is genuinely the only product that does this today on top of your real broker.
When to pick which
Pick M1 if: you want broker + portfolios + cash management in one app, you want scheduled auto-rebalance, you need custodial accounts (UGMA/UTMA), or you're happy to ACATS-transfer your existing assets to M1.
Pick Walnut if: you want to keep your existing broker, you want AI to drive portfolio research and basket-building, you want to ask questions about your live positions in plain English, or you already use Claude / ChatGPT and want them connected to your real portfolio.
Use both if you want set-and-forget on one allocation (M1) and AI-driven exploration on another (Walnut + a different broker like Public or Schwab). Plenty of users do exactly this.
Try Walnut on top of your broker
Walnut works with your existing broker, Public, Alpaca, Schwab, Tradier, Webull for trading; Robinhood, Fidelity, and many more for tracking. Connect in a few clicks.
FAQ
Is Walnut a broker like M1?
+
No. Walnut is an AI layer that sits on top of your existing brokerage account through SnapTrade. M1 is a full broker that custodies your assets and clears trades directly. That's the most important architectural difference between them, and it drives most of the other differences.
Does Walnut have auto-rebalancing like M1's Pies?
+
Walnut shows you the exact trades that would bring your basket back to target weights and lets you execute them through the Invest dialog. M1 lets you set a schedule (weekly, monthly, quarterly) and rebalances automatically. If hands-off is the priority, M1 wins; if you want AI-assisted analysis and review before each rebalance, Walnut.
Can I use Walnut and M1 together?
+
Not directly, M1 isn't on SnapTrade's supported broker list, so Walnut can't read M1 positions today. Many users keep M1 for set-and-forget Pies and use Walnut with a different broker (Public, Schwab, Tradier, etc.) for AI-driven research and thematic basket building. That's a perfectly reasonable two-app stack.
Which has better basket / portfolio building?
+
Different shapes. M1's Pies are great for set-and-forget index-tilted portfolios, visual, drag-and-drop allocation, no AI needed. Walnut's baskets are built in AI conversation: you describe a thesis, the AI proposes constituents and weights, and you approve. Walnut wins if you want exploration; M1 wins if you already know your allocations.
What about Expert Pies vs Walnut's basket templates?
+
M1 curates Expert Pies (target-date, hedge-fund-followers, sector ETFs, etc.) that you can clone. Walnut leans on the AI to generate basket ideas dynamically based on a thesis you describe, more conversational, less template-driven. Both work; the right fit depends on whether you want pre-made or AI-generated starting points.
Which one should I pick?
+
Pick M1 if you want everything in one app, broker + Pies + cash + custodial, and you're happy with set-and-forget rebalancing. Pick Walnut if you want to keep your existing broker, want AI to research and propose basket ideas, and want to talk to your portfolio (via Walnut's built-in chat or Claude / ChatGPT). They're solving overlapping problems differently; many users genuinely use both.
Does Walnut cost anything?
+
No, Walnut is free. You pay nothing for the AI assistant, the basket-building, or the portfolio tracking. M1 is also free for the basics; M1 Plus is $10/mo for margin perks and cash management.
Which has better AI?
+
Walnut, by a lot. M1 doesn't have a conversational AI today, pie construction is manual. Walnut has a built-in agentic AI (Claude Sonnet under the hood) that can call tools to read your portfolio, build baskets, run drift analysis, search the web for current data, and execute through the Invest dialog. It also exposes an MCP connector so Claude Desktop and ChatGPT can talk to your portfolio directly.