Good news, OpenCloud fans – you can once again use your Cloud AI subscription to power the hit, open source, autonomous AI Agentic Harness! But, there is a big flaw in how it is being implemented.
A few hours ago, Anthropic announced via its official developer communications account on X, @clouddevs, that it is changing its cloud paid subscription tiers, introducing a new subtier. "agent sdk" Credit for all paid customers, which they can now allocate exclusively "program related" Uses including external, third-party agents such as OpenGL.
The move is a major reversal from Anthropic’s policy introduced as early as April 2026, which had explicitly barred its AI subscriptions from being used to power such non-Anthropic agents and harnesses, with Anthropic saying they cause capacity and service issues.
The problem was that some cloud customers were paying $20 to $200 per month under Anthropic’s Cloud Pro and Max subscriptions, but were consuming tokens (units of information) worth hundreds, even thousands of dollars above those prices through its OpenClave (and similar autonomous) agents. This was an untenable situation for Anthropic’s finances and its limited compute infrastructure to estimate models for end users.
To be clear, even when it enforced the old prohibition against OpenClave and similar agents last month, Anthropic never completely shut down the ability to use OpenClave for the cloud. Rather, it redirected users to pay through the company’s application programming interface (API), be billed based on usage (priced per million tokens rather than a flat monthly rate as subscriptions offer), or pay for additional usage credits on top of their subscription.
Now, Anthropic Cloud is giving customers another way to use their subscription bill to pay for third-party agents.
However, the restoration comes with a significant catch: programmatic usage is no longer subsidized by the general subscription pool, but is instead limited to a fixed, non-rollover monthly credit, worth $20-$200 depending on your cloud plan, and billed at API rates.
In other words, if you don’t use these new Agent SDK credits, they will expire at the end of the month. And if you use them all, you can’t deduct your normal subscription usage limit to cover any additional usage – you’ll need to purchase additional usage credits instead.
Why did Anthropic block cloud subscriptions from OpenClave (and other third-party agent AI harnesses) in the first place?
To understand why this reinstatement matters, one must look at the technical friction that led to the initial ban on April 4, 2026.
Anthropic’s first-party tools, such as Cloud Code and Cloud Cowork, are engineered to maximize "instant cache hit rates"-A method of reusing already processed text to save expensive computation cycles.
Third-party tools like OpenClave, which allow users to run autonomous agents through external services like Discord or Telegram, were often not optimized for these efficiencies. Boris Cherny, head of Cloud Code, said that these were third-party services "It’s really hard for us to work sustainably" Because they bypassed the caching mechanism that allows Anthropic to offer flat-rate subscriptions.
The sheer volume of data being reprocessed by inefficient agents was endangering the stability of the system for the broader user base. Even with Anthropic’s massive expansion into new hardware—including access 300MW Colossus 1 Data Center And with its 220,000+ GPUs – demand for agentic workflows was outstripping sustainable supply.
New "agent sdk credit" The system solves this technical barrier by shifting the cost of inefficiency to the user. By providing a dedicated dollar-amount credit, Anthropic no longer needs to do this "eat the difference" On unoptimized third-party code. If an agent is inefficient and burns through tokenization, it depletes the user’s new $20 to $200 Agent SDK credit budget faster than it exceeds the value of Anthropic’s fixed monthly subscription tiers.
Anthropic’s new programmatic credit system
The restoration of third party access is broken down into Anthropic’s billing tiers, creating a new hierarchy. "Programmatic power." Here’s how much Anthropic is giving each user in terms of new agent SDK credits (in addition to their normal cloud usage through Anthropic cloud products like Cloud Code, Cloud Cowork, etc.).
| Plan |
Monthly, dedicated agent SDK credits (on top of existing subscription plans) |
usage context |
|
Pro |
$20 |
Use of personalized scripts and lightweight SDKs. |
|
max 5x |
$100 |
Medium agentic automation. |
|
max 20x |
$200 |
Professional-grade development environment. |
|
Team (Premium) |
$100/seat |
Collaborative Team Automation. |
|
Enterprise (Premium) |
$200/seat |
Seat-based high-end enterprise use. |
This system presents a sharp division between "interactive" And "program related" Workflow If you’re chatting with the cloud in the browser or using Cloud Code in the terminal to write code interactively, you’re still benefiting from your standard, high-capacity subscription limit.
As Anthropic technical employee Lydia Haley wrote in a post on X, "To add some clarity: You don’t pay extra. It’s the same subscription, same price per month." Haley has also included the following helpful diagram about how new agent SDK credits work:
However, the moment you use it claude -p For non-interactive tasks like running commands, GitHub actions, or connecting third-party tools like OpenClaw, the system switches to the dedicated agent SDK credits.
Once the Agent SDK credit limit is reached ($20 for Pro plan, $100 for Max 5X, etc.), programmatic usage stops unless the user enables "additional uses" Billing, which is charged at standard, pay-as-you-go API rates.
Importantly, for those who consider the basic subscription model to be an infinite resource, this is a hard limit. Credits do not roll over, which means "Use it or lose it" The nature of the system forces the developer’s budget to be reset monthly.
strategic implications
The licensing implications of this move run deep "agentic" Ecosystem.
By explicitly allowing third-party apps like Conductor and OpenClave to authenticate through the Agent SDK, Anthropic is legitimizing the workflow it previously attempted to block.
However, in doing so it has brought an end to an era "Calculate Arbitrage".In early 2026, a $20 Pro subscription can be leveraged through OpenClaw to run agents that would cost hundreds of dollars for a standard API key.
Moving toward metered credit, Anthropic is aligning its subscription model with its developer platform (API). While it offers a "Free" For customers, Buffer ensures that high-volume, production-level automation is transferred to predictable, token-based billing.
This protects the company’s margins while still offering "sandbox" API-first for developers to experiment without the immediate overhead of an account.
Community reactions are perhaps surprisingly negative
While Anthropic officials framed the update as "simplification"The developer community has largely branded this as a significant reduction in the value of their subscription. The reaction focuses on the sharp disparity between previous effective use and the new, measured reality.
Popular AI YouTuber and T3.gg developer Theo Brown (@theo) warned developers that this change is a massive demotivation for those using external tools. "If you use any of the following with your cloud sub, your usage should be reduced by 25x," Theo said, listing T3 Code, Conductor, Z, and Gene as the affected platforms. He concluded with a sharp warning: "They are disguising it as ‘free credit’. Don’t fall for it".
Kun Chen, a solo builder and former L8 engineer at Meta, Microsoft, and Atlassian, interpreted the move as a complete surrender of Anthropic’s market leadership. "It’s official. Anthropic prohibited all programmatic use of cloud subscriptions," Chen posted that he had found himself "Excitement about OpenAI is increasing rapidly" As a result. Chen argued that "Anthropic’s only edge was on coding, and GPT 5.5 has already overturned this," Indication of possible migration of specific developer talent.
Other builders questioned the practical utility of the proposed credit. Ben Hillack, co-founder and chief technology officer of AI agent observability and governance startup Raindrop.AI, expressed concerns over the sustainability of Anthropic’s infrastructure. "This is either really stupid, or shows how bad the spot anthropic is in re: GPS," Hilack said before explicitly asking users "Guess how many turns the last $20 in API credits takes".
The frustration extended to the marketing of the change. Evernever, creator of Inkstone.UK, expressed disbelief at the policy’s determination. "What are you waiting for?! Find even more ways to use the subscription you’re paying for?! And you dare to make it look like a victory?". The sentiment highlights a growing rift between Anthropic and its power-user base, who feel that previously inclusive features are being canceled under the guise of one. "upgrade."
The bottom line for Anthropic customers and AI builders
anthropic "restoration" This is a strategic move to retain developers while strictly managing the physical limits of computation. till 15th June "agentic" The cloud will be the zeitgeist for customers.
The company has successfully regained control of its margins, even if it has meant losing some of the goodwill of its most vocal power users.
However, for the individual developer or enterprise AI builder relying on the anthropic model for OpenClaw, this is clearly an improvement compared to last month’s complete ban.
<a href