Be wary of Anthropic bans

[benjaminalgreen]

Anthropic explicitly bans unsanctioned third-party use. Unless this project got the blessing of Anthropic, I would expect the OAuth approach to be in violation of their terms. They've been very hawkish at trying to kill projects that take users away from the Claude Code harness (think OpenCode) so I would be very careful.

[jeremymcs]

@benjaminalgreen We all agree on your concern. What you’re pointing at is a bigger shift that’s already starting to play out across the ecosystem.

Right now Anthropic is optimizing for control and safety of their surface area, which makes sense from a risk and monetization standpoint. But historically, platforms that over-constrain access tend to lose momentum to ones that make integration frictionless. Developers don’t just choose the best model, they choose the one that lets them build the fastest, ship the easiest, and own the customer relationship.

The OAuth angle is really about that ownership layer. If a model provider enables clean, user-consented access where the developer can plug into identity, billing, and workflows without fighting the platform, that becomes incredibly attractive. It removes the “walled garden” feeling and turns the model into infrastructure instead of a destination.

If another major model provider leans into that, especially with:
• simple OAuth based user authorization
• clear terms that explicitly allow third party surfaces
• stable API access without fear of sudden enforcement changes
• and pricing that supports embedding into SaaS products

then you’re going to see a pretty fast migration. Not necessarily because the model is dramatically better, but because the developer experience is.

We’ve seen this pattern before. The winners are usually the platforms that maximize distribution, not the ones that try to contain it.

Also, the “vibe” layer you mentioned matters more than people admit. Developers talk. If the sentiment becomes “this platform will shut you down once you get traction,” that alone slows adoption. On the flip side, if a provider is seen as pro-developer and permissive within reason, they become the default choice even if they’re not number one on benchmarks.

The other piece people aren’t saying out loud is switching cost is dropping fast. With abstraction layers, multi-model routing, and frameworks already in place, most teams aren’t truly locked in anymore. If a provider becomes difficult to work with, it’s no longer a six-month rewrite, it’s a config change and some prompt tuning.

So if someone ships:
• OAuth-first access
• explicit blessing of third-party UX layers
• predictable policy enforcement
• and a clear stance that developers can own the customer

then it’s not just “people might switch.” You’ll see entire ecosystems form around that provider almost overnight.

At that point it stops being about model quality alone. It becomes about trust, control, and who’s actually enabling builders versus competing with them.

And if Anthropic doesn’t adjust that posture, they risk becoming a great model that people respect… but don’t build on.

[benjaminalgreen]

I agree. I've been scrambling to try to drop Anthropic since this change. I'm pretty much just waiting for OpenAI to release their $100 pricing tier and/or for GLM-5 to work better.

[timbicker]

As far as I understand, there is a real risk getting banned using claude subscription with the OAuth API. To find out more you can have a look at the recent T3 video which is about T3 Chat. They are using the claude code binary directly which should be fine.

https://youtu.be/RIkSlHgQYog?si=NnWsBai8Fn-2ngjb

[dogc-pi-iac]

i thought anthropic was a PBC. this seems anticompetitive, but i might be misinterpreting.

[jeremymcs]

#3926