Anthropic adds separate Agent SDK credits for paid Claude plans on June 15

• Starting June 15, paid Claude plans get a separate monthly programmatic-usage credit, and ClaudeDevs' announcement says it covers the Agent SDK, claude -p, Claude Code GitHub Actions, and third-party apps built on the Agent SDK.
• According to ClaudeDevs' pricing thread, the new credit ranges from $20 on Pro to $200 on Max 20x, with Team plans getting per-seat credits and Enterprise varying by seat type.
• ClaudeDevs' rollout details say the credit is claimed once, resets each billing cycle, does not roll over, and falls back to optional API-billed usage credits when the monthly bucket runs out.
• Separately, ClaudeDevs' limit update says Claude Code weekly limits are 50% higher through July 13, and that temporary bump stacks with the 2x increase to 5-hour limits announced the week before.
• Reaction split fast: LLMJunky's workaround post said someone had already built a zmux workaround for the claude -p changes, while marckohlbrugge's complaint and petergyang's response focused on product complexity and developer trust.

Anthropic's most useful clarification lives in its support article, which turns a messy rate-limit story into a simple one: chat stays chat, SDK usage gets its own bucket. The other live wrinkle is the weekly-limit bump announcement, which runs through July 13 and stacks with last week's 5-hour increase. And the fastest community response was not a benchmark or a tutorial, but a workaround post plus a small pile of complaints that Claude's packaging has become harder to parse.

The core change is separation. ClaudeDevs says programmatic usage no longer shares the same pool as Claude Code and chat, and Noah Zweben's summary framed it as Agent SDK and -p moving into a new bucket while interactive limits stay unchanged.

That bucket applies to four surfaces:

• Claude Agent SDK, per ClaudeDevs' product list
• claude -p, per the same announcement
• Claude Code GitHub Actions, per ClaudeDevs' coverage list
• Third-party apps built on the Agent SDK, per ClaudeDevs' follow-up

Anthropic's support article is the canonical source linked from the launch thread.

The plan table is simple enough to scan as a list:

• Pro: $20/month, according to ClaudeDevs' plan breakdown
• Max 5x: $100/month, per the same breakdown
• Max 20x: $200/month, per ClaudeDevs' thread
• Team Standard: $20 per seat, per ClaudeDevs' pricing list
• Team Premium: $100 per seat, per the same list
• Enterprise: varies by seat type, according to ClaudeDevs' Enterprise note

ClaudeDevs' usage-flow post says users claim the credit once, after which eligible programmatic calls draw from it automatically. When that bucket empties, usage can continue on metered API-priced credits if the user has turned them on, and pauses until reset if they have not.

The timing is also fixed: ClaudeDevs' rollout schedule says users will get an email on June 8 to claim credits, and the change takes effect June 15.

The second announcement landed the same day. ClaudeDevs says weekly Claude Code limits are already 50% higher for Pro, Max, Team, and seat-based Enterprise users, with the increase live now through July 13.

The thread attached to that post adds four concrete details:

• It applies across the CLI, IDE extensions, desktop, and web
• It requires no opt-in
• It expires July 13 at 6 PM PDT, 1 AM GMT
• It stacks with the 2x increase to 5-hour limits announced the previous week

A small creator-side detail tucked into om_patel5's repost is that /usage in the CLI reflects the updated limits.

The fastest hacker reaction was immediate evasion. LLMJunky said someone had already built a workaround to the claude -p changes using zmux, which turned the announcement into a brief cat-and-mouse story before the new credit system has even gone live.

A different thread of reaction was about trust, not tooling. thekitze's complaint called the move a reputation hit, while marckohlbrugge's post said Anthropic needs to simplify its product lineup and nicklaunches' reply argued that forcing users to think about which option to use can overshadow the product itself.

petergyang used the moment to argue for plainer developer communication, linking to Lee Robinson's essay on building and retaining developer trust. That reaction says as much about Claude's current packaging problem as the credit table does: Anthropic shipped more budget, but the discourse centered on explaining which bucket pays for what.