<h2>What’s changing</h2>
<p>Starting June 15, Anthropic is separating programmatic Claude usage from standard subscription credits. Paid plans will get a dedicated monthly credit pool for agent and automation workloads — $20 for Pro, $100 for Max 5x, $200 for Max 20x, with per-seat amounts for Team plans. This covers <code>claude -p</code>, Claude Code GitHub Actions, Python/TypeScript SDK usage, and third-party apps built on the Agent SDK.</p>
<p>Credits don’t roll over and aren’t transferable. Once they’re gone, you either enable pay-as-you-go API billing or hit a wall until the next cycle.</p>
<h2>Why developers are angry</h2>
<p>The math is brutal for anyone who built workflows around the old flat-rate plans. Previously, tools like T3 Code, Zed, and OpenCode could tap into Claude Code subscriptions via the Agent SDK — usage that Anthropic itself estimated was worth $5,000–$7,500/month in inference. Under the new structure, that same usage is capped at $200/month. That’s a 25–40x reduction.</p>
<p>Theo, who built T3 Code as an open-source Claude UI wrapper using the Agent SDK (the only officially supported path at the time), announced he’s canceling his Anthropic subscription and switching to Codex. His argument: Anthropic explicitly encouraged developers to build on the SDK, then changed the economics underneath them.</p>
<p>This also reverses a policy from just a month ago that had removed third-party tool coverage from subscriptions entirely — so the new credit system is being framed as a partial restoration, but developers who remember what they had before aren’t buying it.</p>
<h2>The longer pattern</h2>
<p>This isn’t a sudden change. Over roughly five months, Anthropic quietly blocked OAuth access for some third-party tools, updated its Terms of Service to formalize the distinction between “interactive” and “programmatic” usage, and is now enforcing it through billing. The official explanation is billing simplification. The skeptical read is that Anthropic wants valuable training-signal interactions routed through its own surfaces — <a href="http://Claude.ai" target="_blank" rel="noopener nofollow">Claude.ai</a> and Claude Code — rather than third-party wrappers, while also improving unit economics ahead of a reported $950B funding round.</p>
<p>A billing bug along the way charged at least one user $200.98 in unexpected API fees after a false-positive detection of third-party tool usage, which didn’t help trust.</p>
<h2>Enterprise implications</h2>
<p>For teams running CI pipelines, autonomous agents, or long-running automations, metered billing complicates budgeting. Credits don’t pool across team members, runaway agents can exhaust a month’s budget in hours, and cost forecasting gets harder. Analysts are advising teams to treat AI agent usage like cloud infrastructure: set hard budget alerts and optimize for token efficiency.</p>
<p>This isn’t unique to Anthropic — OpenAI and GitHub Copilot are moving in the same direction toward consumption-based pricing for agentic workloads. But the speed and reversals in Anthropic’s policy have left developers who built on the platform feeling burned.</p>
<h2>Meanwhile, Salesforce is spending $300 million on Anthropic tokens</h2>
<p>Not everyone is unhappy. Salesforce CEO Marc Benioff says the company expects to spend $300 million on Anthropic tokens in 2026, almost entirely on AI coding agents. He credits AI agents with letting Salesforce cut its support workforce from 9,000 to 5,000. Benioff also teased upcoming coding features inside Slack and said he wants an intelligent routing layer that sends complex tasks to frontier models like Claude while directing simpler ones to cheaper alternatives.</p>
<p>Salesforce has also invested over $300 million in Anthropic directly, holding roughly a 1% stake now worth more than ten times the original investment. Agentforce, Salesforce’s AI product line, hit $800M ARR, up 169% year-on-year.</p>
<p>The contrast is stark: enterprise customers spending hundreds of millions on API access are fine. Developers who built open-source tools on subscription plans are not.</p>


Collections

Anthropic is splitting Claude subscription billing starting June 15, creating a separate monthly credit pool for programmatic/agentic usage ($20–$200/month depending on plan tier). This replaces what was previously unlimited or near-unlimited access via the Agent SDK, representing a 25–40x reduction in effective value for developers who built workflows on top of Claude subscriptions. Developers like Theo (T3 Code) are canceling subscriptions and switching to competitors like Codex, arguing Anthropic encouraged SDK-based development then changed the economics. The shift is part of a broader five-month pattern of Anthropic tightening third-party tool access via OAuth blocks, ToS updates, and now billing separation. Enterprise customers like Salesforce (spending $300M on Anthropic tokens in 2026) are unaffected, highlighting a stark divide between enterprise API customers and indie developers on subscription plans. Teams running CI pipelines or autonomous agents face harder budget forecasting, as credits don't pool across seats and runaway agents can exhaust monthly budgets quickly.

Claude paid plans to include monthly API credit starting June 15

<p>You sure about the Claude Code? Received email yesterday saying:</p>
<p>“Your subscription usage limits don’t change. They stay reserved for interactive usage of Claude Code, Claude Cowork, and chat. See the <a href="https://links.email.claude.com/s/c/PzAeMbLloqJb4BiM7eUby1Ax2Ya1cDNLudy8VyDqX3x0_R0DNoks_cobTpXW-FYXaz6FvyCzIYaumQv97AC7V_Yqiw4UGPuhFIdTAtuusO8Op0FJ67MkVXdaNgM9BTa9DmxblxaO8G14iFPpOCUVovNz6Ydzwqm9taOqEeRuWP6yPt2MRnSsYMUFb41hWJ1SO8fwq5vCLgnHlIRrTmfdTsjf4ZtfyTdFW0BJRbeblQ1lbTeT-OJfUVm6uPBbMJkZAHVVVLmTxSgxxgKNg4cGPl1V884_fLX43Z2O4clDDtYZcCb6MgornRlOslKUsxf_93VHfZqocpWruCgBDYZz5b48sCSDlUFo7WoD1GaVzu4nk8bnBhHw8lBxhTlgwPZBtWNqHPsK7HeSRRySpdicF-xRGEh0mZ_x2qI_iXkeVXugdgFsDXGPhpy76NN6o0pHzCoJIAypMyfSF2FgSrIg2Jo6dduEQyPNKB17zLIsFZMRiwDa1STrdqhk8nbQwKzf1gYaWYTJOHBaKkhTxeXIU6E1YFZwerX53IpcQHTmc9LW8kIKNa3DSeaHIUd_8MFBZkNXE_ZGyVT_Vip9HV75TgDpaSrRxtC7gfvhH87_tHoyk147nm4RaFbevfLsENX8B6MXssQG86Py76UsC9zwO1tIe_mSbhuNuPN_6ELFQjypSfU5ZMKIZW0S9kAc1WhIsb2_oPn72T-VuyNMJF63Kzx6VkOX8xMpFjHfhZfc7iqvg6QM7-zqb-zLhXdd39KjDfk4DaWmPHsen28oztgKsmA-Hm_dDd5j6DZlZ52f-jRV0g/qjkmQbpjiDP6tiG6VwQsHn4D0oHb5BRE/20" target="_blank" rel="noopener nofollow">Help Center</a> for more details.”</p>
