Is there this but for local electron clients? I want to use Claude pro/max with MCP and openai plus and gemini etc (even cursor/roo/cline) via the terminal without paying extra for the api as that makes no sense. Preferably one that has tabs to actually use multiple at the same time.
This is quite excellent though; makes it easier to try different solutions and switch between them (considering they are all doing a race to get better).
It’s not too bad to get this working through dev tools and playwright, but I’m hesitant to bundle it into a repo or share my elisp because I don’t want to get banned by Anthropic.
In my opinion pretty terrible that github nilly willy can decide your org was inactive for too long and give it a new owner. (tbf i dont really know for how long and how inactive but still, this is potentially a security vulnerability with github)
This looks to me like a great way to log claude code interactions.
Is it just me that cannot figure out a way to log the prompts and responses when using Claude Code?
If I look here I see nothing about a logging switch (beyond --verbose which is different). I'm surprised I cannot tell it to write this to a per session file.
It is interesting because my daughter uses my claude account on another computer and I can see her prompts and she can see mine. I get nervous when I go down a political rabbit hole that she will get freaked out, but so far so good. Claude does keep a "history" there. But, not when you are using the code tool.
There is one very good reason I can imagine why this would not be in the tool. If I'm billing a client at $100X and my final cost from Claude Code is only $X, then I suppose I would prefer that no one could ever ask for that information. Someday employers will, I bet.
Maybe there is an easier way, like scripting or capturing the terminal output directly?
I haven't used claude-task-master before, but based on the README, it looks like it's an AI agent that integrates well with IDEs. In contrast, AgentAPI lets you control other agents - like Claude Code or OpenAI Codex - using HTTP calls instead of typing commands into the terminal. For example, you could use AgentAPI to control Claude Code from a custom frontend, such as a native desktop application.
This is cool. I've been using claude code over SSH which also works very well. Even was using it via my home vpn on my phone while on a walk.
Seems like you could achieve the same type of experience with this.
Is there this but for local electron clients? I want to use Claude pro/max with MCP and openai plus and gemini etc (even cursor/roo/cline) via the terminal without paying extra for the api as that makes no sense. Preferably one that has tabs to actually use multiple at the same time.
This is quite excellent though; makes it easier to try different solutions and switch between them (considering they are all doing a race to get better).
It’s not too bad to get this working through dev tools and playwright, but I’m hesitant to bundle it into a repo or share my elisp because I don’t want to get banned by Anthropic.
Here’s a hint: https://github.com/microsoft/playwright/issues/10369#issueco...
How did you get that org name!?
It was inactive for a long time [1]. I emailed GitHub about it and they gave it to us!
[1]: https://web.archive.org/web/20201108093051/https://github.co...
Wow.
For all the usernames I've ever tried getting (and failed cause they don't give out inactive usernames on most platforms), that's amazing.
In my opinion pretty terrible that github nilly willy can decide your org was inactive for too long and give it a new owner. (tbf i dont really know for how long and how inactive but still, this is potentially a security vulnerability with github)
Probably since no one has abused this mechanism yet. I’m sure if people start trying to vacuum up inactive accounts, it might become a problem
Can you share the e-mail?
This is pretty cool! If there's an MPC server we can set up Cursor and Claude Code to peer program on a project:)
I'm really fascinated by this.
This looks to me like a great way to log claude code interactions.
Is it just me that cannot figure out a way to log the prompts and responses when using Claude Code?
If I look here I see nothing about a logging switch (beyond --verbose which is different). I'm surprised I cannot tell it to write this to a per session file.
https://docs.anthropic.com/en/docs/agents-and-tools/claude-c...
It is interesting because my daughter uses my claude account on another computer and I can see her prompts and she can see mine. I get nervous when I go down a political rabbit hole that she will get freaked out, but so far so good. Claude does keep a "history" there. But, not when you are using the code tool.
There is one very good reason I can imagine why this would not be in the tool. If I'm billing a client at $100X and my final cost from Claude Code is only $X, then I suppose I would prefer that no one could ever ask for that information. Someday employers will, I bet.
Maybe there is an easier way, like scripting or capturing the terminal output directly?
Was just looking for something like this in the past few days
can you compare this to https://github.com/eyaltoledano/claude-task-master ?
I haven't used claude-task-master before, but based on the README, it looks like it's an AI agent that integrates well with IDEs. In contrast, AgentAPI lets you control other agents - like Claude Code or OpenAI Codex - using HTTP calls instead of typing commands into the terminal. For example, you could use AgentAPI to control Claude Code from a custom frontend, such as a native desktop application.
This is a developer. Task manager is a project manager.