Anthropic is removing third-party harness access from Pro subscriptions effective immediately. Thousands of users are discovering this morning that their $20/month plan no longer covers their favorite tools.
No heads-up. No transition period. No discussion. One email, and it’s done.
Starting April 4, 2026 at 12pm PT / 8pm BST, your Claude Pro subscription no longer covers usage through OpenClaw — or any other third-party harness. If you want to keep using them, you’ll need to enable “extra usage,” a pay-as-you-go billing on top of your monthly subscription.
You were already paying. Now you’ll pay again.
Your Claude Pro subscription ($20/month) no longer covers usage limits through third-party harnesses including OpenClaw. To keep using them with your Claude account, you must enable “extra usage” — billed separately. Claude Code and Claude Cowork remain covered.
Anthropic’s email: word for word
Here’s what Anthropic sent to Pro subscribers this morning:
“Starting April 4 at 12pm PT / 8pm BST, you’ll no longer be able to use your Claude subscription limits for third-party harnesses including OpenClaw. You can still use them with your Claude account, but they will require extra usage, a pay-as-you-go option billed separately from your subscription.”
The wording is clever. Technically, you can “still use” third-party tools. Just not with what you’re already paying for. A feature became a surcharge, with no change to the base price.
Claude Code and Cowork are spared — for now
A key detail from the email: “Your subscription still covers all Claude products, including Claude Code and Claude Cowork.”
Anthropic is drawing a clear line between its own products and the rest of the ecosystem. If it’s made by Anthropic, it’s covered. If it’s third-party, it costs extra. The message is crystal clear: the open ecosystem is only tolerated if it pays its way.
Claude Code
The autonomous AI dev agent, right in your terminal
Claude Cowork
The AI-powered collaborative workspace by Anthropic
”Managing capacity”: the excuse that doesn’t hold up
Anthropic justifies this move by citing the strain third-party tools put on their infrastructure. Their exact words:
“We’ve been working to manage demand across the board, but these tools put an outsized strain on our systems. Capacity is a resource we manage carefully and we need to prioritize our customers using our core products.”
Here’s the thing: a token is a token, whether it’s sent from claude.ai or from OpenClaw. GPUs don’t check the passport of the request. What bothers Anthropic isn’t “system strain” — it’s that these users actually consume the resources included in their subscription.
The key phrase: “we need to prioritize our customers using our core products.” Translation: if you use Claude outside our apps, you’re not really our customer anymore.
The tools directly impacted
OpenClaw
The open source AI agent that turns your LLMs into autonomous workers
OpenClaw is named explicitly as the first tool affected. But the rest is clear:
“This policy applies to all third-party harnesses and will be rolled out to more shortly.”
Every third-party tool that accesses Claude through subscriber credentials is now on borrowed time. The open ecosystem Anthropic was promoting just months ago — “build on Claude” — has been walled off from the inside.
The “gift”: a one-time credit, 13 days to spend it
To compensate for a permanent business model change, Anthropic offers a one-time credit:
“To make the transition easier, we’re offering a one-time credit for extra usage equal to your monthly subscription price. Redeem your credit by April 17.”
A credit worth roughly $20, expiring April 17. Thirteen days. After that, every request outside the walled garden gets billed at full rate.
Better yet: the company is offering “discounts up to 30%” on pre-purchased extra usage bundles. Translation: we’re removing a feature from your plan and offering to sell it back to you in bulk at a discount. Bold move, at least they own it.
If you depend on third-party Claude access, you can: enable extra usage (pay-as-you-go on top of your sub), switch to the Claude API directly (separate pricing), or explore alternatives: GPT-4o, Gemini 2.5, DeepSeek V3.
Zero notice: the method that hurts
The email arrives in the morning. The cutoff hits the same evening. And Anthropic promises a second email the next day — the one offering a refund to those who want out.
“You will receive another email from us tomorrow where you’ll have the ability to refund your subscription if you prefer.”
The next day. After the damage is done.
In any other industry, unilaterally changing the terms of a paid service with a few hours’ notice would be called unfair business practice. In AI tech, they call it a “policy update” and put a “Claim your credit” button underneath.
What this says about Anthropic in 2026
The company built its identity on safety, transparency, and ethics — the deliberate counterpoint to OpenAI. This decision tells a different story: when GPUs are expensive and investors are watching margins, values become negotiable.
The developers and power users who evangelized Claude when the general public had never heard of it just learned where they really stand: cost line item, margin adjustment variable.
Key takeaways
- Cutoff date: April 4, 2026 at 12pm PT / 8pm BST
- Tools affected: OpenClaw first, all third-party harnesses next
- Still covered: Claude.ai, Claude Code, Claude Cowork
- Compensation: one-time credit = 1 month subscription, must be redeemed by April 17
- Discounts: up to 30% off pre-purchased extra usage bundles
- Refund: refund email promised for the next day (April 5)
Anthropic is offering refunds. But what’s broken can’t be refunded.
Trust doesn’t come with an extra usage credit.
Claude
The AI that understands nuance, by Anthropic
Source
Official email sent by Anthropic to Claude Pro subscribers on April 4, 2026:
Tools mentioned
Find the best AI tool
108 tools compared, tested and rated. Find the one that fits your needs.