-
Notifications
You must be signed in to change notification settings - Fork 624
Description
Problem
The Agent SDK requires API key authentication and explicitly prohibits using claude.ai subscription billing (Pro/Max). This creates a significant barrier for individual professionals and small teams building autonomous agent systems.
The workflow looks like this:
- Developer pays for Max ($100-200/mo) to use Claude Code interactively
- Developer builds automation on top of Claude Code (exactly what the SDK is for)
- The moment they call the SDK programmatically instead of typing in a terminal, they're forced onto a separate usage-based billing system with no spending cap
This is a penalty for building the kind of autonomous systems Anthropic is actively promoting.
Impact on ecosystem adoption
I'm building an open source agent coordination system that orchestrates Claude Code workers across repositories. My users — and my customers' users — often already have Pro or Max subscriptions. When I tell them the automated pipeline requires a separate API account with pay-per-token billing, the reaction is consistently: "I'm already paying for this, why do I need to pay twice?"
I'd rather onboard my customers directly to Anthropic subscriptions. Many of them already have Pro accounts. The current model forces me to explain a billing split that feels arbitrary from their perspective — especially when claude -p (the CLI) does work with Max billing, but the SDK (which does the same thing programmatically) does not.
Suggestion
Allow the Agent SDK to authenticate against Max plan billing, perhaps with a configurable concurrency or token budget that fits within the plan's existing limits. The CLI already supports this — the SDK should too.
Even a limited version (e.g., Max plan with lower concurrent agent cap) would remove the adoption barrier for individual professionals who want to automate their workflows without managing a separate billing relationship.