主张在 OpenClaw 部署中使用 Anthropic API 以获取高级模型，并通过上下文管理优化 Token 用量。

taking the other side of this OpenClaw debate around switching from Claude to other models now that Anthropic is forcing API usage 

my openclaw @charles__AI 🦀 was built from the start using Anthropic API so we could access Opus, Sonnet, Haiku and for me it’s been worth it. 

operates seamlessly without much problems. built some great memory and messaging systems. 

yes it costs more but you can optimize and save on tokens in many ways and if you’re trying to build a serious project, like I am with @charles__AI, you should invest in the best infrastructure to do so

seeing too many horror stories already of people switching to chatGPT to save some money 

open to hearing from others though who’ve also been using the API on ways to further optimize on token usage

optimizing on unnecessary context injections has been a big one for me