Meta put a price on Meta One and rolled it out worldwide: 6.99 euros a month for the Core plan and 20.99 for Premium, with the paid features of Instagram, Facebook and WhatsApp inside the same bundle, plus expanded Meta AI quotas, according to Hipertextual's coverage. In Mexico, Core costs 119 pesos and Premium 399. La Nación adds that Instagram Plus and WhatsApp Plus had already been running in Argentina at 2 and 1.5 dollars a month. The same day, and with far less press, a WhatsApp Business MCP server for coding agents showed up.
Why it matters
The headline is the subscription. The signal is something else: the Core plan comes with 500 image generations and 200 video generations per month. That's not a feature, that's a meter. Meta went from selling attention — something that scales for free with advertising — to selling compute, which does not scale for free and therefore gets counted. The company itself says it plainly: the apps and Meta AI remain free, and what you pay for is "expanded AI usage, which requires more computing capacity".
Translated for anyone building software on top of these platforms: the free tier is frozen at whatever it already does, and everything new that depends on a GPU is born with a quota. If your product lives off features Meta could bundle, take a good look at that list of more than 50 features: several of them are things somebody once paid a third-party tool for.
What changes in practice
First, the arithmetic. The three individual plans added together come to 8.47 euros; Core costs 6.99. The bundle isn't a pricier option with perks: it's the economical default, and it's designed that way.
| Plan | What it includes | EUR/month | MXN/month |
|---|---|---|---|
| WhatsApp Plus | WhatsApp paid features | 2.49 | 29 |
| Instagram Plus | Instagram paid features | 2.99 | 39 |
| Facebook Plus | Facebook paid features | 2.99 | 39 |
| Meta One Core | All three, plus 500 images and 200 videos | 6.99 | 119 |
| Meta One Premium | Core with expanded AI limits (no figures published) | 20.99 | 399 |
Second, and for me the most relevant thing of the day, is the WhatsApp Business MCP server. If you've ever integrated that API you know what I'm talking about: registering the number, verification, message templates that get rejected without telling you why, the test environment, webhooks that never arrive. It's tedious work, badly documented and hard to hand off to someone junior. That's exactly the kind of task where an agent with console context pays off: it doesn't write your product, it takes the paperwork off your plate.
Careful with the enthusiasm: for now, all we have on the MCP server is the announcement and TechCrunch's summary. I don't have the documentation on scopes, permissions or test environment in front of me, so any claim about what that agent can touch in your account would be made up. What I do know from the general MCP pattern is what you need to check before connecting it: which credentials it receives, whether it distinguishes a test account from a production account, and whether it can send real messages or only draft them.
When NOT to use it
Don't build a product's cost model on the Premium plan. Core publishes its numbers; Premium says "more limits" without saying how many. An input with no declared unit doesn't belong in a cost model. If you need predictable volume, an API with per-unit pricing is healthier than a consumption subscription, even if the monthly ticket looks higher.
Don't buy it as a team tool yet. Meta mentions four plans for businesses and creators, but the published material doesn't detail how they're billed, whether there's centralized administration, or how access is transferred when someone leaves. If you manage client accounts, that missing information is the problem, not the price.
Don't connect the MCP to a client's production account as your first experiment. An agent with permissions over templates and messaging can, worst case, message real people. Test account, your own number, and only then the conversation with the client.
And don't ignore the other side of things getting cheaper. Generating convincing audio and video costs less and less, and voice-cloning fraud on WhatsApp already has concrete victims, as this La Nación piece documents. I'm not saying Meta One causes that: I'm saying that if you build messaging products, out-of-band identity verification stopped being an optional roadmap feature.
What I'd do today
With the MCP, one afternoon: test account, a throwaway number, and I ask the agent to put together three templates and submit them for approval. If that works, I save hours on every client onboarding and it's worth documenting the flow. With Meta One, nothing: 6.99 euros doesn't move the needle for anyone, and the decision that matters isn't whether I subscribe, it's not resting any piece of product on features that can move between plans without warning. The meter is already in place. Better to design assuming they're going to use it.