Loading
← Blog
2 min read

I Don't Like Anthropic, and It Has Nothing to Do With the Model

When I code, I feel more comfortable in Pi or OpenCode than in Claude Code. The problem is I can't use a Claude model there without paying for it separately through the API. To me, that's a mistake, not a reasonable decision.

  • AI
  • Developer Tools
  • Anthropic
Developer working at a cluttered desk at night, surrounded by monitors showing code and terminal windows

Most of my coding happens in Pi or OpenCode, not Claude Code. I feel more at home there: they're lighter, they let me pick a model depending on the task, and I don't have to carry everything Claude Code brings with it. The problem is that if I want to use a Claude model there, I have to pay for it through the API, on top of the subscription I already pay, and it costs a lot more than using it directly inside Claude Code.

To me that's a mistake, not a reasonable decision. It makes no sense that my subscription works fine inside Claude Code and stops being useful the moment I open the tool I actually prefer.

What bothers me most is that there's no technical excuse behind it. Anthropic is the company that built the Model Context Protocol, the open standard that Claude Code, Pi, OpenCode, and basically every AI tool use to connect to outside data. At the protocol level, everything works fine between these tools. The problem was never that they don't integrate well. It's that Anthropic decided my subscription only counts inside its own tool.

And this isn't me preferring Claude Code and putting up with the rest. It's the opposite. I prefer working in Pi or OpenCode. I use Pi when I want something minimal, without all the weight of Claude Code, and I use OpenCode when I want to choose between different models depending on the task. By choosing the tool I feel more comfortable in, I end up paying extra to use the same model that comes included with my subscription inside Claude Code.

I get that it doesn't suit Anthropic for third-party tools to use their subscription-tier model without them seeing that money as API revenue. That's a fair business reason. But they solved it the worst way possible: instead of competing to make Claude Code the tool I'd actually prefer, they blocked me from using the same model in the tool I do prefer without paying extra.

In the end, my problem with Anthropic was never how good the model is. It's that the company behind it decided to treat my preference for a different tool as a threat to its business, instead of competing to earn that preference.