Anthropic Hates Developers | Josh Symonds

A Tale of Submitting to Two App Directories Savecraft connector review, 2026 OpenAI<br>March Submitted<br>TOOT! ✓ April Approved

Anthropic<br>March Submitted twice<br>April Customer support escalation to nowhere<br>May "In the standard process"<br>June "In our review queue"<br>NOPE July Application lost

I wasted four months waiting for Anthropic to review my MCP for its Connector Directory. Its support bot promised a human response that never came, while real humans from enterprise support and Platform Operations explicitly assured me that my application was proceeding normally. It was not, despite $800 per month in personal spending, a six-figure enterprise relationship I helped drive, and open-source tooling I maintain for Claude Code. This is part of a broader pattern of silence and ineffective support, including a serious Claude Code bug that drops roughly 23 percent of Fable’s user-facing messages.

Claude is extraordinary. The company surrounding it is quite the opposite. Receipts follow.

Anthropic hates developers.

I do not mean that its employees wake up each morning wondering how to make programmers miserable. I mean that institutions reveal their priorities through what they fund, what they measure, and what failures they are willing to tolerate.

Anthropic appears to value developers when we consume tokens, improve benchmark scores, advocate for Claude, and build free infrastructure around its products. The moment we need reliability, transparency, support, or basic reciprocity, Anthropic disappears.

Let me explain:

I maintain four Claude Max 20x accounts personally, representing $800 per month in personal spending. At work, my advocacy helped drive a substantial six-figure annual relationship with Anthropic spanning employee adoption and API usage.

I am quite confident that this does not make me Anthropic’s largest customer. It should, however, be enough to establish that I am not casually kicking the tires. In almost any other vendor relationship involving this much money, I would expect a support channel, a recognizable escalation path, and some minimally coherent service expectations.

Instead, my requests are lost, shuffled between departments, answered inaccurately, and often simply ignored. And my experience outside of requests is one of frustration, absent ownership and accountability, and degraded service.

Anthropic’s commitment to developers seems to begin and end with releasing increasingly capable models into the world. I love better models — Claude is an extraordinary piece of software, and it is why I have spent so much money and time here! But the experience of depending on an AI platform does not end at model intelligence: it includes availability, tooling, documentation, support, billing, quotas, communication, ecosystem stewardship, and the ability to report a serious defect without arguing with a chatbot that insists the defect does not exist.

And everything surrounding Claude is so consistently hostile and alienating that “Anthropic hates developers” has become the most parsimonious explanation for me.

How can I say this? I am so glad you asked.

Four Months in a Queue Anthropic Abandoned

I created an application called Savecraft. I am proud of it!

I am not shilling Savecraft here, but a discussion of its basic capabilities is germane, so please bear with me.

Savecraft connects games to AI assistants. It can read save files, retrieve character and inventory data, consult game documentation, and give an assistant enough context to provide useful advice without requiring the player to manually transcribe everything into a chat window.

Savecraft is an MCP server at its core. When Anthropic and OpenAI announced connector directories, I was excited about publishing it through both. I especially liked the possibility of building rich interfaces that could appear directly inside AI conversations (and I think I succeeded materially, to be un-humble).

I submitted Savecraft to Anthropic’s connector directory on March 22, eight days before I submitted it to OpenAI. At the time, Anthropic’s submission mechanism was a Google Form. There was no dashboard, status page, or reliable confirmation that the application had reached the correct place.

On March 30, after a substantial expansion of Savecraft’s UI and capabilities, I filled out the entire Anthropic form again so the reviewers would have current information. That same day, I submitted Savecraft to OpenAI’s connector directory.

OpenAI approved it on April 28: 29 days from submission to acceptance. Since then, it has attracted meaningful traffic through ChatGPT, and I have developed a responsive relationship with OpenAI’s platform team around subsequent reviews and improvements.

My experience with Anthropic was… rather different.

When I opened a support request to ask whether Anthropic had received it, its automated support system suggested that I submit the form again. And so, I did.

Then I...