xenodium
Claude Code Max Plan OAUTH authentication
Claude Code has been included in the Claude Max Plan subscription. This bypasses the ANTHROPIC_API_KEY authentication. Also with a Max Plan it seems that you cannot create any API-Tokens for claude code anymore. You have to use oath for authentication.
For claude code integration with chatgpt-shell we would need an additional oauth mechanism in chatgpt-shell.
just have asked the Anthropic "Support" if API-Keys will be supported for claude code in MAX Plan, but bot says NO :)
Catching up here... is the feature request to enable chatgpt-shell to send requests and show output from Claude code? While I'm not yet familiar with Claude Code APIs or integration points, but from chatgpt-shell perspective we would likely need a new/different service handler from what we currently have for Anthropic.
Do you have examples of how to integrate or what APIs are offered? Are you keen to offer a PR?
I am very new to both Claude AI and AI in general, and also from an Emacs perspective really more of a user than a developer. I just thought it would be neat to integrate Emacs with Claude AI and your project is what I found...
I tried to install it and got stuck at the "set key as emacs variable" part since I can't find a key.
In the settings there are several options to connect Claude AI with something called an MCP server: https://support.anthropic.com/en/articles/11175166-getting-started-with-custom-connectors-using-remote-mcp
Not sure this is what would be needed?
I just thought it would be neat to integrate Emacs with Claude AI and your project is what I found...
I hear ya. Keep in mind it's mostly just me behind this project, so unless I can sustainably work on this project, it's hard for me to keep up with new AI capabilities. It's quite a lot of work.
I tried to install it and got stuck at the "set key as emacs variable" part since I can't find a key.
Most cloud LLM services require an API key to set clients up (including chatgpt-shell). Alternatively, you may try local services which don't require an API key. For example, ollama.
In the settings there are several options to connect Claude AI with something called an MCP server: https://support.anthropic.com/en/articles/11175166-getting-started-with-custom-connectors-using-remote-mcp
chatgpt-shell currently acts as a client connecting to LLM models which act as the servers. The MCP integration would require a reverse logic (chatgpt-shell becomes a servier). This is non-trivial work at the moment.
Not sure this is what would be needed?
Thanks for this. While it describes how it works, I still requires non-trivial amount of work to integrate it. Also, I'm not quite sure if we'd be able to present the claude code output as a native chat shell.
Thank you for your extensive reply! I will continue to monitor this issue and can test stuff if something gets written.