Anthropic announced that Claude has resumed support for OpenClaw, using "Agent SDK credits" to curb subscription arbitrage.

📄Full Article· Automatically extracted by trafilaturaGemini 翻譯2551 words

Anthropic announced the reopening of subscription access to third-party agent tools (such as OpenClaw), while simultaneously introducing a separate "Agent SDK Credits" mechanism—the monthly allowance expires when used up and does not roll over, ending the "arbitrage window" where users could run hundreds of dollars worth of compute on a $20 subscription. (Recap: Meta launches AI "end-to-end encrypted" incognito chat, Zuckerberg: even we can't see it) (Background: From TON to BIO: Why is the market starting to repriced strong-narrative assets?) Anthropic announced this morning (14th) on its official developer account @ClaudeDevs: paid subscription users will receive a brand-new "Agent SDK Credits" sub-allowance, specifically usable for third-party autonomous agent tools including OpenClaw (a directional policy reversal, as the company had previously blocked such tools). Starting June 15, paid Claude plans can claim a dedicated monthly credit for programmatic usage. The credit covers usage of: – Claude Agent SDK – claude -p – Claude Code GitHub Actions – Third-party apps built on the Agent SDK— ClaudeDevs (@ClaudeDevs) May 13, 2026 On the surface, this looks like a concession to the developer community; but in substance, Anthropic has completed a redrawing of its financial boundaries: users can still use third-party agents, but the cost of all programmatic calls will no longer be borne by the subscription fee pool. In early April this year, Anthropic explicitly prohibited subscription users from using Claude with non-Anthropic-proprietary agent frameworks such as OpenClaw. The official reason was "capacity and service stability," but the real driver was financial logic: some users paid fixed subscription fees of $20 to $200 per month, yet consumed hundreds or even thousands of dollars' worth of Tokens through autonomous agents. Claude Code lead Boris Cherny pointed out the core contradiction: Anthropic's own tools are deliberately optimized for "prompt cache hit rates"—reusing previously processed text fragments as much as possible to reduce redundant computation costs. Third-party tools like OpenClaw bypass this caching mechanism, meaning nearly every conversation round requires recomputation, which is "hard to make sustainable" for Anthropic. Even though Anthropic has secured over 220,000 GPUs worth of compute resources at the Colossus 1 data center, demand growth for agent workflows still outpaces what the infrastructure can absorb. The current lifting of the ban comes because Anthropic has found a technical solution to push costs back onto the user side. According to Anthropic's announcement on X, the monthly Agent SDK credit caps for each plan are as follows: - Pro: $20 / month - Max 5x: $100 / month - Max 20x: $200 / month - Team (Premium): $100 / seat / month - Enterprise (Premium): $200 / seat / month When users execute the claude -p command, trigger a GitHub Action, or connect to third-party tools such as OpenClaw, the system automatically switches to the Agent SDK credit pool for billing, at the same rate as the API (priced per million Tokens). Once credits are exhausted, programmatic usage is suspended; any remaining balance at month-end does not carry over. Anthropic technical staffer Lydia Hallie added on X: To add some clarity: you don't pay extra. It's the same subscription, same price per month. What's new our sub now covers two separate pools: · Interactive → sub limits, unchanged · Programmatic → new $20–$200 included(!!) credit, metered at API rates https://t.co/zdd2PsE60p pic.twitter.com/knrcT1P5Of— Lydia Hallie ✨ (@lydiahallie) May 13, 2026 The most immediate impact of this policy is ending the "compute arbitrage" phenomenon that emerged in early 2026: using $20 Pro subscription privileges to run autonomous agent workflows via OpenClaw equivalent to hundreds of dollars in standard API costs. T3.gg founder Theo Browne warned developers: "If you use any of the following with your Claude sub, your usage may have been cut by 25x," and stated directly: If you use any of the following with your Claude sub, your usage must got cut by 25x: – T3 Code – Conductor – zed – jean – "Claude -p" in your ci – scripts to call Claude code from other toolsThey're disguising this as "free credits". Don't fall for it. https://t.co/5LMGkUrp7z — Theo – t3.gg (@theo) May 13, 2026 Atlassian senior engineer Kun Chen said bluntly: "Anthropic is officially pulling the plug on all programmatic subscription usage," and admitted he is becoming increasingly confident in OpenAI. Raindrop.ai CTO Ben Hylak's comment was sharper: "Either this is extremely stupid, or it shows how bad Anthropic's situation is on GPU resources." Explicitly allowing tools such as Conductor and OpenClaw to authenticate through the Agent SDK effectively endorses the third-party ecosystem; but by designing an independent credit pool, Anthropic is also passing the marginal costs of inefficiency back to end users, rather than having them absorbed by the subscription fee pool. (One can expect that most related competitors will pivot to the same strategy in the future—after all, compute remains a scarce resource for now.)

Data Status✓ Full text extractedRead Original (動區 BlockTempo)

🔍Historical Similar Events· Keyword + Asset Matching6 items

2026-05-25

The White House has reached an agreement with Anthropic, and the NSA will fully integrate Claude AI.

2026-05-24

OpenClaw Core Contributors: After the Lobster craze fades, who should Agents listen to?

Similarity 170%關鍵字 openclaw/agent同分類 zh

2026-05-20

A Deep Dive into the Story Behind Anthropic's Ban Wave: The Safety Religion, the AI Civil War, and Claude's Dilemma Amid US-China Decoupling

2026-05-20

Claude Managed Agents opens self-hosted sandboxes and MCP encrypted channels; Anthropic is bringing the execution environment into enterprise intranets

2026-05-19