Read-Only: What Toast Won't Let You Build

In July 2025 I asked Toast how to pull my own sales numbers out of my own register. My account rep answered the next day: what I wanted was the Analytics API, and getting it meant buying their Restaurant Management Suite Pro tier, at $150 a month per location.

I have three locations. Reading my own numbers: $450 a month.

I wrote back that night, and I’d send it again today. “I think $450/mo for management suite pro is crazy.” Then the line I still chew on: “This along with being able to copy menu items between locations were both free with square.”

So I asked for the smaller version. Not a partnership, not a product, just a narrower tier: read-only, API-only, a trial to prove it was worth something. All I was trying to do was stop spending hours a week hand-exporting my own sales history into spreadsheets.

Then I bought the $450 tier anyway.

That’s the part that earns me the right to write the rest of this. I’m not the operator who got quoted a number, got mad, and walked. I paid it. It’s been on my bill ever since, and I’d pay it again, because hours a week of copying my own numbers into a spreadsheet by hand is worse.

What I want you to understand is what the money bought. It bought reading.

An API, if you’ve never had a reason to care, is the plug on the side of a piece of software: the place where another program can connect and either ask questions or give instructions. Every modern system has one. The rest of this post is about which direction Toast lets the plug work, and what each direction costs.

Asking questions costs $450 a month. Giving instructions isn’t sold.

Toast’s own developer documentation describes what I bought in exactly those terms: “Your integration can only read data from Toast.” A smaller detail on that page stuck with me harder. It doesn’t come with a sandbox, the practice environment developers use to test without touching anything real. You pay to look, and looking is all you get.

Write access is a different door with no price on it. The orders API, the one that would actually put an order in my register, accepts writes from “partner integrations and custom integrations” only. Partners are the DoorDashes and the online-ordering companies, with a signed agreement, a certification, a program.

A month after I bought in, I emailed Toast’s developer support asking how to get into their Developer Portal. The answer was polite, fast, and clearer than anything on the pricing page. The Developer Portal, they explained, “is a feature to enable partner developers,” and since I’m “a customer developer using our Standard and Analytics APIs through your regular Toast interface,” I have no need for one.

Read that twice. There are partner developers and there are customer developers, and Toast will tell you which one you are. The difference between the two isn’t skill or intent or even money, because I’d already paid. It’s which side of the register you’re standing on. Customer developers read. Partner developers build.

I’ll grant the obvious, same as last time: Toast is good software, I run it every day, and some gate on order injection has to exist. An open order endpoint with no vetting would flood kitchens with garbage inside a week. But look at what the gate is made of. A safety inspection has a standard you can meet. A partner program has a business development team.

The phone that can’t take an order

Now the part that made me write this. The clearest thing AI can do for a restaurant in 2026 is answer the phone. Not a “press 2 for hours” menu: a voice that hears “two large pepperonis, one with jalapeños,” answers the questions, and books the order while your counter person keeps the line moving. The technology for that is done. I’ve already built an AI that handles phone calls for my own errands: it plans the call, waits for my OK, dials, and logs what happened. Pointing the same trick at the shop’s inbound line is a weekend project, not a moonshot.

Except the last step. When the caller finishes ordering, the robot has to put the order somewhere, and “somewhere” means writing into the register, and I’m not a partner integration. So the robot can hear the order perfectly and then hold it like a note it isn’t allowed to pass. Best case, it reads the order back to an employee who retypes it into Toast by hand. I’d have automated the easy half of the job and kept all the labor.

And before anyone decides I’m just doing it wrong: Slang.ai is probably the best-known phone AI in restaurants, at $399 a month per location for the Core plan and $599 for Premium. Their own pricing page lists what it connects to: OpenTable, SevenRooms, Yelp. Reservation systems. Not one tier on that page writes an order into a POS. The wall I hit as one owner with a laptop is the same wall a funded software company builds its pricing around. The robots that CAN put an order in your Toast got there the only way anyone gets there, through the partner program, and you rent them as subscriptions at partner prices.

The ordering page you rent

Second locked door, same lock. I pay Toast $133.06 a month for my own online ordering page. The invoice line calls it “Digital Storefront Pro.” Suppose I’d rather own that page. My domain, my menu, my regulars, no checkout I don’t control. Building the page is the easy part now: the site you’re reading came out of the same tools that run my back office. But an ordering page has exactly one feature that matters, moving the order into the register, and that’s the same partner-gated write as the phone robot.

So here is the real menu for an independent. Rent Toast’s page at $133.06 a month. Rent ChowNow’s at $249 to $449 a month plus a setup fee. Rent Owner.com’s at $499 a month, about a thousand dollars up front, plus 5% of every order your own customers place on your own website. Or push everybody to the delivery apps and lose the margin there instead. There is no door marked “build it yourself.”

The app they’ll sell you instead

You can’t build your own app either, and by now the reason is boring: an app is a menu, a card form, and the ability to place an order, and ingredient three is the locked one. Ship it anyway and you’ve made a brochure with an icon.

Toast knows the app is worth money, because Toast sells one. In October 2024 they launched branded apps, your name and your colors on their order flow, sitting in the top Digital Storefront tier at $299 a month. I pay $133.06 for the tier below it. The pitch of the $299 product is, more or less, “looks like you built it.”

Line the three doors up and read the incentive out loud. The phone robot, the ordering page, the app: each one is something an owner could build on a register that took instructions, and each one is revenue as long as the register doesn’t. The page and the app, Toast sells outright. The order-taking phone robots arrive through Toast’s partner program. Opening write access to owners wouldn’t cost Toast server capacity. It would cost them the catalog. None of this needs a villain, and I’m not telling a villain story. A pricing team doing its job gets there all on its own. But you should know whose job the lock is doing while you pay rent on the door.

Why I’m building Tempo, part two

The $2,000-a-month post gave the first reason I’m building my own register: $2,016.73 a month in Toast software subscriptions across my restaurant and two trucks, $24,200.76 a year, on a bill with no ending built into it. I audited my way to 18% off one invoice and the trim was the whole prize. That was the arithmetic reason.

This post is the other reason, and it’s the bigger one. The subscription rents you the counter. It does not let you build on it, and paying more doesn’t change that, which I know because I paid more. For $24,200.76 a year I get a register that talks fluently to DoorDash, to Grubhub, to the certified partners of Toast, Inc., and goes quiet when its own owner asks it to take one instruction from something he built himself. Every idea I have for my own shop routes through somebody else’s partner program and somebody else’s roadmap.

Tempo is the same counter with the locks off. A register, a kitchen screen, online ordering wired straight into both, and a dashboard that reads the day the way I read it: sales by the hour, next to labor. And the part this post exists to say: when I want Tempo to do something new, there is no application process and nobody to ask. The register takes instructions from its owner. That sentence is the entire product.

Which brings me to the money, and I’ll handle it the way this site handles every number: real ones or none.

Tempo is not switched over yet. It rings test orders in the back office, and its kitchen screen has already run a Friday night quietly next to the Toast one. So today’s ledger is two columns, and only one of them is a savings number.

The column I can prove: $2,016.73 a month of Toast software subscriptions, which becomes a hosting bill the day the switch flips. The read access this whole post is about is a line item inside that number, not a bill sitting next to it. I’ve promised to publish the replacement bill next to the invoices it replaces, and I will.

The second column never shows up anywhere, and it’s the one that changed how I think. It’s the price of the things I couldn’t build. A phone that answers itself: $399 a month per location, from the best-known vendor, and it still can’t finish the order. My own ordering page and app: $499 a month plus 5% of my own customers’ orders. I don’t pay a dollar of that today, so none of it goes on a P&L as savings and I won’t pretend otherwise. But it’s the going rate for capabilities my own register could have handed me, and I already bought the right to read.

What to do about yours

You should almost certainly not build your own POS. That advice from last time stands. But if you’re signing with any register company, or renewing, get two answers in writing before the ink:

• What does API access to my own data cost, and is it read-only?
• Who is allowed to write an order into my register, and what would it take for that to be me?

The rep will have a crisp number for the first one. Mine did, inside a day. Listen to how long the second answer takes, because that silence is the whole product roadmap. Then pull up what you already pay for the capabilities this post priced, put those subscriptions in one column, and add the ones you talked yourself out of buying. That column is what the lock costs you, and you pay it either way: in money if you buy the door, in hours if you don’t.

I bought the door. It opens one direction. So I’m building my own, and when the switch flips you’ll get the number.