To understand the fight between Anthropic and the developer community that uses Claude Code, you need to know how the business worked before the conflict started.
The trick developers were using
Claude Code is Anthropic's official AI coding tool for the terminal. It has two payment models: a monthly subscription (starting at $20/month on the Pro plan, up to $200 on Max) with unlimited usage within certain limits, and a pay-per-token API where each generated token has a price. The cost difference between the two is enormous: what costs more than $1,000 per month on the API for intensive use costs $200 on the subscription.
What many developers discovered is that third-party tools — open-source projects like OpenCode and Claudebot — could connect to Anthropic's servers using the same OAuth subscription credentials. Put simply: you could install OpenCode, log in with your $200/month Claude Max account, and use Claude with the same power as the official tool — but with a different, more flexible interface, without paying API rates.
For Anthropic, this was a direct economic problem: they were unwillingly subsidizing thousands of users consuming tokens at subscription prices when they should have been paying per token. For developers, it was the most accessible way to use Claude in their workflows.
The first move: forcing a name change
The first public confrontation was more symbolic than economic. Peter Steinberger had created a project called Claudebot — an open-source tool for using Claude from the terminal with advanced features. Anthropic contacted Steinberger and demanded he change the project's name, arguing that using "Claude" in a third-party product's name could cause confusion and violated brand policies.
Steinberger changed the name: Claudebot became OpenClaw. The decision didn't sit well with the community. Many saw it as the first symptom that Anthropic was going to harden its stance against the independent tools ecosystem built on Claude. Steinberger, according to The New Stack, ended up joining OpenAI shortly after.
The January block: OpenCode gets cut off
In late January 2026, Anthropic implemented technical measures to block third-party tools from accessing Claude through subscription OAuth credentials. Anthropic engineer Thariq Shihipar explained the move on social media: third-party tools using Claude through subscriptions generated "unusual traffic patterns without any of the telemetry that the official Claude Code client provides," making it difficult to debug and manage the network.
The impact was immediate. OpenCode — one of the most popular alternative clients, with thousands of users — stopped working with Claude credentials. In its GitHub repository, the OpenCode team published a commit removing support for Claude Pro and Max account keys. The commit message was direct: "anthropic legal requests."
xAI — Elon Musk's AI lab — also lost Claude access through this mechanism. The block reached any tool that used Claude Code's internal OAuth system for authentication.
The reaction on Hacker News, Reddit and X was one of fury. Developers argued that Anthropic was destroying the open-source ecosystem that had grown around Claude, that subscriptions had no explicit clause prohibiting that use, and that the company was prioritizing revenue over the community that had helped it grow.
The February legal clarification
In February 2026, Anthropic updated its legal documentation to make explicit what had previously been implicit. The new policy stated that OAuth tokens obtained through Claude Free, Pro and Max accounts "are only intended for Claude Code and Claude.ai" and that using them in "any other product, tool or service — including the Agent SDK — is not permitted and constitutes a violation of the Consumer Terms of Service."
Anthropic clarified that the update "represents an attempt to clarify existing policy language to make it consistent throughout company documentation" — meaning the rule always existed, it was now just written explicitly.
Tools using the official paid API key, Amazon Bedrock or Google Vertex AI were not affected. The block only impacted those using the subscription OAuth system to access Claude from unofficial tools.
Ten days later: the source code leak
Context matters here. On March 31, 2026 — exactly ten days after Anthropic sent cease-and-desist letters to OpenCode — the Claude Code source code leak happened on npm. The more than 600,000 lines of TypeScript that were exposed showed, among other things, that Claude Code has in development an autonomous daemon mode (KAIROS) that works very similarly to what OpenClaw was already doing: an agent that operates in the background without the user initiating it manually.
The irony wasn't lost on the community: Anthropic was blocking third-party tools for doing something that its own official tool already had built and waiting to launch behind a feature flag.
The leak reversed community sentiment in less than 48 hours. The same developers who days earlier had written that "Anthropic is a gatekeeping megacorp" were now fascinated, analyzing the leaked code's architecture.
The current state: OpenClaw survives, with conditions
In early April 2026, Anthropic announced that starting April 4 at noon, Claude subscriptions would no longer cover usage in third-party tools like OpenClaw — with an important caveat: usage would still be possible through a system of "extra usage bundles" available at a discount, billed separately from the subscription.
OpenClaw's official documentation reflects the current state: the project team states that "Anthropic staff told us this usage is allowed again" under certain conditions, and that they treat Claude CLI access as authorized unless Anthropic publishes a new policy changing that.
claw-code — a clean-room rewrite of Claude Code's architecture in Rust and Python created after the source code leak — remains available on GitHub with more than 75,000 stars. Anthropic retracted most of its DMCA orders on April 1, focusing legal action only on repositories containing the directly leaked source code, not on rewrites.
What the conflict revealed
The fight between Anthropic and the independent tools ecosystem exposed a tension that goes beyond a single case: the AI lab business model faces a structural conflict between the open-source community that helps spread and improve their tools and the need to control access so that revenue remains sustainable.
Claude Max at $200/month is cheaper than the API token equivalent because Anthropic assumes subscription users won't run the model at maximum capacity all the time. When a third-party tool automates coding loops 24 hours a day, that assumption breaks — and Anthropic ends up subsidizing intensive use at casual-use prices.
The question left open in the developer community is whether this will be the pattern for all major AI labs, or whether one of them will find a way to monetize without closing the ecosystem that made them popular.
Comments
💬 Log in to comment💬 Join the conversation and log in to comment.