Anthropic Got Caught
An Anthropic employee posted this in March: “i bought a mac mini so i could have blue bubbles when texting claude and it started roasting me… try the imessage plugin for claude code today.”
That’s an employee at Anthropic publicly promoting a plugin that reverse-engineers Apple’s iMessage infrastructure. Their official plugin directory. Install instructions and screenshots included.
It breaks Apple’s terms of service. Apple says you can’t reverse-engineer their software, access their services through automated scripts, or use iMessage for commercial purposes. This isn’t a grey area.
About a week earlier, Anthropic’s lawyers had forced Open Code to rip out their Claude integration. Same month. You can’t make this up.
What Open Code Actually Did
Open Code built a plugin that let paying Claude subscribers use their $200/month subscription through a different interface. Same API. Same account. Same subscription you’re already paying for. Just a different frontend.
Anthropic killed it. They updated their consumer terms on February 19, 2026, to explicitly ban the use of subscription OAuth tokens in third-party tools. The same day, Open Code’s maintainers pushed a commit titled “anthropic legal requests” and gutted all Claude OAuth code from the repo.
Their terms say it plainly: “Using OAuth tokens obtained through [Claude Code/Claude.ai] in any other product, tool, or service, including the Agentic SDK, is not permitted and constitutes a violation of the consumer terms of service.”
Your subscription. Their approved CLI. No exceptions.
Now Read Apple’s Terms Again
Apple says: we built iMessage. It works on our devices, through our app, using our servers. You agreed not to reverse-engineer it or access our services through unauthorized tools.
Anthropic says: we built Claude Code. It works through our CLI, using our subscription endpoint. You agreed not to use your OAuth tokens outside our approved interfaces.
That’s the same argument. Word for word, principle for principle. Access our service however you want, as long as it’s through the interface we control.
Anthropic just happens to be on both sides of it. Enforcing it against developers who build competing interfaces. Violating it when it’s someone else’s platform.
The Receipts
Open Code was forced to remove their Claude OAuth integration after legal pressure. February 2026.
JetBrains faced similar restrictions. Their AI Assistant can access Claude models, but only through JetBrains’ own separate subscription. Not your personal Claude Pro or Max plan. Anthropic won’t allow it.
Matt Pocock said it best: “I have never before experienced, from any developer tool, such a frustrating lack of clarity over the basic terms of usage.”
The terms have been updated multiple times since launch. Developers building in good faith can’t get a straight answer on what’s allowed.
That’s not a bug. Ambiguity lets you enforce selectively.
Apple Looks More Open. Apple.
Here’s the part that should embarrass them. In this comparison, Apple comes out looking like the permissive one. Apple. The walled garden people. The 30% platform tax people.
Don’t like iMessage? Install WhatsApp. Install Telegram. Same phone. Works fine. Apple doesn’t charge you 50x more for using a competitor’s messaging app.
Don’t like Claude Code’s CLI? Pay per-token through the usage-based API at dramatically higher rates. You can use other interfaces. It’ll just cost you way more.
Pick One
I don’t think Anthropic’s restrictions should be illegal. Companies should be able to define the terms of their products. Apple shouldn’t be forced to open-source iMessage. Anthropic shouldn’t be forced to let every third-party tool tap into their subscription endpoint.
But you can’t send legal threats to developers for routing your subscription through unauthorized interfaces and then turn around and ship a product that does the exact same thing to Apple’s platform. In the same month.
Either access restrictions matter. Or they don’t. Pick one.