Good news, OpenClaw fans: you can once again use your Claude AI subscription to power the successful, open-source, autonomous AI harness! But there is a big problem with the way it is implemented.
a few hours ago, Anthropic announced through your official developer communications account at X, @ClaudeDevswho is changing his Claude paid subscription levels, introducing a new subcategory of "Agent SDK" credits for all paid subscribers, which can now be assigned specifically for "programmatic" uses, including third-party external agents such as OpenClaw.
The measure represents a significant setback with respect to Anthropo’s strategy. policy introduced in early April 2026 which expressly prohibited its AI subscriptions from being used to power these types of non-Anthropic agents and harnesses, after Anthropic said they caused capacity and service issues.
The problem was that some Claude subscribers were paying between $20 and $200 per month under Anthropic’s Claude Pro and Max subscriptions, but consuming hundreds, even thousands of dollars in tokens (units of information) above those prices through their OpenClaw agents (and similar freelancers). This was an unsustainable position for Anthropic’s finances and its limited computing infrastructure to infer models to end users.
To be clear, even when it enacted the old ban against OpenClaw and similar agents last month, Anthropic never completely cut off Claude’s ability to be used in OpenClaw. Rather, it redirected users to pay through the company’s website. application programming interface (API)which is billed per usage (priced per million tokens, rather than a flat monthly fee like subscriptions offer), or you pay for additional usage credits on top of your subscriptions.
Now, Anthropic is offering Claude subscribers another way to use their subscription bill to pay third-party agents.
However, the rollout comes with a major drawback: programmatic usage is no longer subsidized by the general subscription pool, but is instead restricted to a fixed, non-transferable monthly credit, which is also worth between $20 and $200 depending on your Claude plan, and is billed based on API rates.
In other words, if you don’t end up using these new Agent SDK credits, they simply expire at the end of the month. And if you use them all, you can’t fall back on your general subscription usage limits to cover any additional usage; instead, you will need to purchase additional usage credits.
Why did Anthropic block Claude’s subscriptions from OpenClaw (and other third-party AI harnesses) in the first place?
To understand why this restoration is important, you have to look at the technical friction that led to the initial ban on April 4, 2026.
Anthropic’s own tools, such as Claude Code and Claude Cowork, are designed to maximize "fast cache hit rates"—a method of reusing previously processed text to save on costly computing cycles.
Third-party tools like OpenClaw, which allow users to run autonomous agents through external services like Discord or Telegram, were often not optimized for these efficiencies. Boris Cherny, director of Claude Code, noted that these third-party services were "It’s really difficult for us to do it sustainably." because they overlooked the caching mechanisms that allow Anthropic to offer flat-rate subscriptions.
The large volume of data reprocessed by inefficient agents threatened the stability of the system for a broader user base. Even with Anthropic’s massive expansion into new hardware, including access to Colossus 1 300 MW Data Center and its more than 220,000 GPUs: demand for agent workflows was outstripping sustainable supply.
the new "Agent SDK Credit" The system solves this technical bottleneck by transferring the cost of inefficiency to the user. By providing exclusive dollar credit, Anthropic no longer has to "eat the difference" in non-optimized third-party code. If an agent is inefficient and consumes tokens, it simply depletes the agent’s new SDK credit budget $20 to $200 faster, rather than exceeding the value of Anthropic’s fixed monthly subscription levels.
Anthropic’s new programmatic credit system
Third-party access restoration is segmented into Anthropic billing levels, creating a new hierarchy of "programmatic power." This is what Anthropic gives each user in terms of new Agent SDK credits (in addition to their normal use of Claude through Anthropic Claude products like Claude Code, Claude Cowork, etc.).
|
Plan |
Monthly Dedicated Agent SDK Credit (in addition to existing subscription plans) |
Context of use |
|
Pro |
$20 |
Individual scripts and light SDK usage. |
|
Maximum 5x |
$100 |
Moderate agent automation. |
|
Maximum 20x |
$200 |
Professional level development environments. |
|
Equipment (Premium) |
$100/seat |
Automation of collaborative teams. |
|
Company (Premium) |
$200/seat |
Large-scale seat-based business use. |
This system introduces a marked division between "interactive" and "programmatic" workflows. If you’re chatting with Claude in a browser or using Claude Code in a terminal to write code interactively, you’re still taking advantage of the limits of your standard high-capacity subscription.
As Lydia Hallie, a member of Anthropic’s technical staff, wrote in a post on X, "To add some clarity: you don’t pay more. It’s the same subscription, the same price per month." Hallie also included the following helpful diagram of how the new Agent SDK credits work:
However, when you use the claude -p command for non-interactive tasks, running a GitHub action, or connecting a third-party tool like OpenClaw, the system switches to the dedicated agent SDK credit.
Once the Agent SDK credit limit is exhausted ($20 for Pro plans, $100 for Max 5X, etc.), programmatic usage stops unless the user has enabled "extra use" billing, which is charged at standard pay-as-you-go API rates.
Crucially, for those who found the original subscription model to be an infinite resource, this is a hard limit. Credits do not accumulate, which means that "use it or lose it" The nature of the system forces a monthly reset of the developer’s budget.
Strategic implications
The implications of this licensing measure are profound for the "agent" ecosystem.
By explicitly allowing third-party apps like Conductor and OpenClaw to authenticate through the Agent SDK, Anthropic is legitimizing a workflow it had previously attempted to block.
However, in doing so, it has ended the era of "calculate arbitrage"By early 2026, a $20 Pro subscription through OpenClaw could be leveraged to run agents that would cost hundreds of dollars with a standard API key.
By moving to metered credit, Anthropic is aligning its subscription model with its Developer Platform (API). Although it offers a "free" buffer for subscribers, ensures that high-volume production-level automation moves to predictable token-based billing.
This protects the company’s margins while offering a "sandbox" for developers to experiment without the immediate overhead of an API-first account.
Community reactions may be, unsurprisingly, negative
While Anthropic executives framed the update as a "simplification"the developer community has largely called this a significant reduction in the value of their subscriptions. The reaction centers on the marked disparity between the previous actual use and the new measured reality.
Popular YouTuber and AI developer Theo Browne (@theo) from T3.gg warned developers that this change constitutes a massive devaluation for those using external tools. "If you use any of the following with your Claude sub, their usage should be reduced by 25x," Theo stated, listing T3 Code, Conductor, Zed, and Jean as affected platforms. He concluded with a stern warning: "They are disguising this as “free credits”. Don’t fall for it".
Kun-chena solo builder and former L8 engineer at Meta, Microsoft and Atlassian, interpreted the move as a complete relinquishment of Anthropic’s leadership in the market. "It’s official. Anthropic disconnected ALL programmatic use of Claude’s subscription," Chen posted and added that he had found "increasingly optimistic about OpenAI" as a result. Chen argued that "The only advantage of Anthropic was encoding, and gpt 5.5 has already changed that," indicating a possible migration of elite developer talent.
Other builders questioned the practical usefulness of the credits offered. Ben Hylakco-founder and CTO of AI agent governance and observability startup Raindrop.ai, expressed concern about the sustainability of Anthropic’s infrastructure. "This is really dumb or it shows how bad the anthropic place is in re: gpus," Hylak noted: before bluntly asking users to "Guess how many spins are worth $20 in API credits?".
The frustration extended to the marketing of the change. NeverNevercreator of inkstone.uk, expressed disbelief at the formulation of the policy. "Wait what?! Are you taking away more ways for me to use the subscription I’m paying for? And you dare to make it look like a victory?". This sentiment highlights a growing divide between Anthropic and its power user base, who feel that once-inclusive features are being rescinded under the guise of a "improvement."
The bottom line for Anthropic subscribers and AI creators
anthropic "restoration" It’s a tactical move to retain developers while strictly managing the physical limits of computing. By June 15, the "agent" The era for Claude subscribers will be metered use.
The company has successfully regained control of its margins, even if it has cost them some of the goodwill of its most prominent users.
However, for the individual developer or enterprise AI creator relying on Anthropic models for OpenClaw, it is clearly an improvement over last month’s blanket ban.
