Wait we can't use claude -p around other tools? What is the point of the JSON SDK then? Anthropic is confusing here, ugh.

edit: And specifically i'm making an IDE, and trying to get ClaudeCode into it. I frankly have no clue when Claude usage is simply part of an IDE and "okay" and when it becomes a third party harness..

▲

cortesoft 7 hours ago | parent | next [-]

I was pretty sure that claude -p would always be fine, but I looked at the TOS and it is a bit unclear.

It says in the prohibited use section:

> Except when you are accessing our Services via an Anthropic API Key or where we otherwise explicitly permit it, to access the Services through automated or non-human means, whether through a bot, script, or otherwise.

So it seems like using a harness or your own tools to call claude -p is fine, AS LONG AS A HUMAN TRIGGERS IT. They don’t want you using the subscription to automate things calling claude -p… unless you do it through their automation tools I guess? But what if you use their automation tool to call your harness that calls claude -p? I don’t actually know. Does it matter if your tool loops to call claude -p? Or if your automation just makes repeated calls to a routine that uses your harness to make one claude -p call?

It is not nearly as clear as I thought 10 minutes ago.

Edit: Well, I was just checking my usage page and noticed the new 'Daily included routine runs' section, where it says you get 15 free routine runs with your subscription (at least with my max one), and then it switches to extra usage after that. So I guess that answers some of the questions... by using their routine functionality they are able to limit your automation potential (at least somewhat) in terms of maxing out your subscription usage.

	▲	ElFitz an hour ago \| parent [-]
		15? Per month? What’s even the point?

▲

hmokiguess 7 hours ago | parent | prev | next [-]

Wouldn't ACP be better for an IDE? https://agentclientprotocol.com/get-started/introduction

	▲	unshavedyak 7 hours ago \| parent [-]
		Possibly, though at first i was entirely focusing (and still am) on Claude Code usage. Given that CC had an API, i figured its own SDK would update faster/better/etc to new Claude features that Anthropic introduces. I'm sure ACP is a flexible protocol, but nonetheless i was just aiming for direct Claude integration.. and you know, it's an official SDK, seemed quite logical to me. It would be absurd to me if the same application is somehow allowed via ACP but not via official SDK. Though perhaps the official SDK offers data/features that they don't want you to use for certain scenarios? If that were they case though it would be nice if they actually published a per-SDK-API restrictions list. That we're having to guess at this feels painful. edit: Hah, hilariously you're still using the SDK even if you use ACP, since Claude doesn't have ACP support i believe? https://github.com/agentclientprotocol/claude-agent-acp

▲

grafmax 7 hours ago | parent | prev [-]

They’re shooting themselves in the foot with these dumb restrictions.

▲

taytus 7 hours ago | parent [-]

They are not dumb restrictions. They just don't have the compute. That is the dumb part. Dario did not secure the compute they need so now they are obviously struggling.

▲

joshstrange 7 hours ago | parent | next [-]

The restrictions are dumb not because they're lower than any of us want them to be, but because they're unclear. Every time Claude comes up on Hacker News, someone asks this question. And every time people chime in to agree that they also are unclear or someone weighs in saying, no, it's totally clear, while proceeding not to point at any official resource and/or to "explain" the rules in a that is incompatible with official documentation.

Example: https://news.ycombinator.com/item?id=47737924

▲

stavros 3 hours ago | parent | next [-]

There's another part that's bullshit: If you've paid for an annual subscription, for a given number of tokens, welp, now you're getting fewer tokens. They've decreased the limits mid-subscription. How is it not bait-and-switch to pay for something for a year only to have something else delivered?

	▲	mcmcmc an hour ago \| parent [-]
		It is, but good luck getting the FTC to care. Maybe the EU will do something about it.

▲

taytus 6 hours ago | parent | prev [-]

You are arguing something different. My point is that they must apply these restrictions. Do I think they could have calculated their growth a little better? Yes, of course, but hindsight is 20/20.

	▲	joshstrange 5 hours ago \| parent [-]
		We might be talking past each other, I promise I'm not just trying to argue. > My point is that they must apply these restrictions. I fully understand and respect they need restrictions on how you can use your subscription (or any of their offerings). My issue is not there there _are_ restrictions but that the restrictions themselves are unclear which leads to people being unsure where the line is (that they are trying not to cross). Put simply: At what point is `claude -p` usage not allowed on a subscription: - Running `claude -p` from the CLI? - Running `claude -p` on a Cron? - Running `claude -p` as a response to some external event? (GH action, webhook, etc?) - Running `claude -p` when I receive a Telegram/Discord/etc message (from myself)? Different people will draw the line in different places and Anthropic is not forthcoming about what is or is not allowed. Essentially, there is a spectrum between "Running claude by hand on the command line" and "OpenClaw" [0] and we don't know where they draw the line. Because of that, and because the banning process is draconian and final with no appeals, it leads to a lot of frustration. [0] I do not use OpenClaw nor am I arguing it should be allowed on the subscription. It would be nice if it was but I'm not saying it should be. I'm just saying that OpenClaw clearly is _not_ allowed but `claude -p` wouldn't be usable at all with a subscription if it was completely banned so what can it (safely) be used for?

▲

mcmcmc an hour ago | parent | prev | next [-]

It’s dumb to piss off their customers with confusing rule changes instead of just raising their prices to deal with high demand. They might even make a profit

▲

dgellow 7 hours ago | parent | prev [-]

Their growth over the past months has been more than insane. It’s completely expected they don’t have the compute. You don’t have infinite data centers around

	▲	taytus 6 hours ago \| parent [-]
		Like or not, openai isn't having the same compute strain, meaning this was predictable.