One of the technical things I don't really understand, is that it is not that hard to have an agent use tmux/etc to type into Claude code directly if you really want to? Like using tmux vs Claude -p are effectively the same aren't they?
Like, I understand the way they are subsidizing or whatever but on a technical level, I don't see how this makes any meaningful difference?
Am I missing something? Is this more about policy than the specific way it is invoked?
There is no difference between claude -p and typing into their "harness" in tmux. They want to make you pay for API which is subsidizing these plans.
They say they want to do it due to their percieved misuse(via automation) of `claude -p`, this I find laughable because they can always rate limit based on how fast a human types/STT's into a text box. Heck, they could provide TUI based extensions(something you vibe code in a few days today) to get better observability and DX for harness use, I would not mind using `claude` CLI directly, but they wont.
Forcing people to using their garbage harness — 0 observability, no extensibility and poor DX — is a showcase of how this fiasco is playing out: They refuse to improve their CLI and want to push everyone to use their API with multiple poorly rolled out policies.
So ive been using claude -p from last year with a harness around it called codelayer(i am sure some of you have been using something similar too). Their harness is completely garbage when it comes at displaying full details of tool calls and session management. A example is how pressing ctrl-o does not display any detail of messages above the scroll bar buffer. It also flashes and rescrolls like bonkers.
The codelayer harness is great at exposing to me thinking tools and tool calls by scrolling and clicking and expanding it. I can scroll and click into any tool call and search the entire session with / less style even if the string is buried in a subagent or tool call. Entire session timeline are visible on the right if I need to navigate to a user/assistant message. None of this is in the default harness.
This isn’t great but with how much they subsidize the max subscriptions it’s also not shocking.
On the bright side, at least we finally have some clear communication around using the agents SDK. I wonder if this in turn will cause more people to build custom agents on top of the SDK or if more people will move off it completely.
I find anyone trying to strike it rich by building on top of the cloud AIs to just be degenerate gamblers. There's every indication that none of these services will be reliable the same way google was for a decade. They're going to squeeze the cash out of you as soon as they think they can.
On the plus side it is fun to use agents with my subscription, on the downside i really really like using Claude Code via ACP is Zed and now this is nerfed.
These seem like not the same thing and it seems silly to bill them the same way but differently than using the CLI or official app.
I'm not sure I understand the intention here-
One of the technical things I don't really understand, is that it is not that hard to have an agent use tmux/etc to type into Claude code directly if you really want to? Like using tmux vs Claude -p are effectively the same aren't they?
Like, I understand the way they are subsidizing or whatever but on a technical level, I don't see how this makes any meaningful difference?
Am I missing something? Is this more about policy than the specific way it is invoked?
There is no difference between claude -p and typing into their "harness" in tmux. They want to make you pay for API which is subsidizing these plans.
They say they want to do it due to their percieved misuse(via automation) of `claude -p`, this I find laughable because they can always rate limit based on how fast a human types/STT's into a text box. Heck, they could provide TUI based extensions(something you vibe code in a few days today) to get better observability and DX for harness use, I would not mind using `claude` CLI directly, but they wont.
Forcing people to using their garbage harness — 0 observability, no extensibility and poor DX — is a showcase of how this fiasco is playing out: They refuse to improve their CLI and want to push everyone to use their API with multiple poorly rolled out policies.
So ive been using claude -p from last year with a harness around it called codelayer(i am sure some of you have been using something similar too). Their harness is completely garbage when it comes at displaying full details of tool calls and session management. A example is how pressing ctrl-o does not display any detail of messages above the scroll bar buffer. It also flashes and rescrolls like bonkers.
The codelayer harness is great at exposing to me thinking tools and tool calls by scrolling and clicking and expanding it. I can scroll and click into any tool call and search the entire session with / less style even if the string is buried in a subagent or tool call. Entire session timeline are visible on the right if I need to navigate to a user/assistant message. None of this is in the default harness.
Damn. I just built an entire headless automated workflow around `claude -p`
I did too. did you do this mainly for observability reasons and a better harness?
This isn’t great but with how much they subsidize the max subscriptions it’s also not shocking.
On the bright side, at least we finally have some clear communication around using the agents SDK. I wonder if this in turn will cause more people to build custom agents on top of the SDK or if more people will move off it completely.
I find anyone trying to strike it rich by building on top of the cloud AIs to just be degenerate gamblers. There's every indication that none of these services will be reliable the same way google was for a decade. They're going to squeeze the cash out of you as soon as they think they can.
And google is one of these services so
On the plus side it is fun to use agents with my subscription, on the downside i really really like using Claude Code via ACP is Zed and now this is nerfed.
These seem like not the same thing and it seems silly to bill them the same way but differently than using the CLI or official app.
Currently:
The credit doesn’t apply to:
Interactive Claude Code in the terminal or IDE
Claude conversations on the web, desktop, or mobile apps
Claude Cowork
Other features that draw from extra usage
------------
Next month, interactive claude code will no longer be allowed.
You can no longer use your Claude subscription with apps that use the Claude Agent SDK, e.g. claude -p, Conductor, T3code, etc.
Yet another developer hostile decision.
(Not that I don't blame them)
[dead]
[dead]