Anthropic Pauses Credit Change for Claude Code – Agent SDK Still on Subscription

Anthropic has paused the planned change that would have moved the Claude Agent SDK, claude -p, and third-party apps from subscription rate limits to a dedicated monthly credit. In an email to users, the company stated the change will not take effect today, and they are working on a revised plan to better support how users build with Claude subscriptions.
What This Means for Developers
Nothing changes right now. According to the email:
- Agent SDK – continues to draw from your existing subscription rate limits.
claude -p– pipes and stdin usage remains covered under the same subscription.- Third-party apps built on the Agent SDK – also continue as before, no separate credit required.
Users do not need to claim any credits, and subscription limits are unchanged. Anthropic says they will provide advance notice before any new policy takes effect.
Community Reaction on HN
On Hacker News, users raised a few points:
- clarity on “works as it did before” – Some question whether this explicitly condones heavy usage via
claude -pand the Agent SDK, since prior policy was ambiguous about potential rate limiting or bans. - no explanation for the reversal – The email provides “zero context” on why the change was pulled, leaving speculation about internal pushback or technical hurdles.
For now, developers relying on the Agent SDK or claude -p for automated or batch use cases can continue without worrying about new credit consumption. Keep an eye on Anthropic's announcements for the revised plan.
📖 Read the full source: HN AI Agents
👀 See Also

Claude Code source leak reveals autoDream memory system and multi-agent patterns
Anthropic accidentally shipped Claude Code's TypeScript source in npm source maps, revealing autoDream memory consolidation, modular system prompt architecture, and multi-agent coordinator patterns.

ETH Zurich Study Questions Value of AGENTS.md Files for AI Coding Agents
New research from ETH Zurich finds LLM-generated AGENTS.md files reduce AI agent task success by 3% and increase inference costs by over 20%, while human-written files offer only marginal 4% gains with similar cost increases.

Leaked Claude Code CLI source reveals hidden features and internal flags
Analysis of leaked Claude Code CLI TypeScript source code reveals 35 build-time feature flags, including BUDDY AI pets, KAIROS persistent memory, ULTRAPLAN remote planning, and Coordinator Mode. Also found 120+ undocumented environment variables and 26 internal slash commands.

Analysis of 413K AI Agent Runs Reveals What Makes Them Succeed
An analysis of 413,278 AI software engineering agent runs from the CoderForge-Preview dataset shows that human software engineering best practices often harm agent performance. The data reveals specific patterns that separate successful from failing runs on the same problems.