Anthropic announced on Wednesday a change that affects every Claude subscriber: starting June 15, there will be a separate ‘Programmatic Credit Pool’ for AI agent usage. Interactive use (chatting, Cowork, documents) and programmatic use (Agent SDK, Claude Code headless, third-party tools like OpenClaw) will be billed separately.
What’s changing
Every paid Claude subscription gets a monthly credit pool matching the subscription price: Pro users get $20, Max 5x gets $100, Max 20x gets $200. These credits apply exclusively to programmatic access — anything running through the Agent SDK, headless mode, or tools like OpenClaw.
Important detail: programmatic usage is billed at API rates, which are significantly higher than the flat subscription rate. If you exhaust your pool, you can keep going — but you’ll pay per token on top. And unused credits don’t roll over.
The backstory
This change has a bumpy history. Just weeks ago, Anthropic cut off third-party agent access through subscription accounts entirely. Boris Cherny, Head of Claude Code, cited server strain from OpenClaw usage. The community was not happy.
Then Anthropic briefly tested removing Claude Code from the Pro plan — and quickly reversed course after headlines hit. The new credit pool is the third attempt at balancing programmatic and interactive usage.
What this means
For casual users, nothing really changes. If you use Claude occasionally for chat, you won’t notice a difference. But for power users running agents around the clock, it’s getting more expensive. $20 in API credits burns fast when an agent is actively ‘thinking.’
Anthropic isn’t alone here: GitHub is moving Copilot to usage-based billing on June 1. The industry is still searching for the right pricing model for agents that never sleep. The flat subscription doesn’t work when an agent consumes in one night what a human uses in a month.
Sources: SiliconANGLE, VentureBeat, The New Stack