Anthropic bans third party auth
no more token of appreciation.
Today Anthropic (creators of Claude) banned the use of their OAuth tokens with third party tools.
This means that you cannot use your Claude plan in other non-Claude apps.
This is of little surprise after they stopped supporting OpenCode (a terminal code agent) around a month ago. There are a few reasons for this:
Inference costs are eating margins:
Third-party applications allow power users to automate workflows and run loops (like Ralph Wiggum) that maximise token usage far beyond what normally would be possible. Anthropic effectively loses money on these power users, and thus limiting their plans to their own products allows them to mitigate against this.
Ecosystem lock-in
By killing third-party integrations, Anthropic forces users into their proprietary environments (like Claude Code). Users who setup and customise Claude Code are less likely to switch in the future, and any friction helps a lot where usually switching AI providers is as simple as a single line of code.High quality training data
When users interact through a thousand different wrappers and custom setups, the data Anthropic collects becomes noisy and impossible to standardise. By funneling all usage into a native environment like Claude Code, they can fine-tune future model to be perfectly tuned to the product. They can optimise the model to execute perfectly within the environment they control, which can be seen in updates like today where improvements to their web search were announced.
Overall I’m not a fan of this move, I’ve long thought that Anthropic artificially boost the value of their plans by having high API costs. This makes sense from their perspective as a business, but it’s unfortunate as a consumer. Luckily OpenAI (creators of ChatGPT) openly encourages using their OAuth tokens in third party apps at the time of writing!
Hope you guys enjoyed/learned something new! Let me know if there’s anything you want me to write about, and don’t forget to like and subscribe!!