Scale · founder · 6 min read
Anthropic Turned Claude Into an App Store. Here's What That Means for You.
Claude Marketplace launched with 2,000+ connectors, agents and consulting partners — a discovery win for Anthropic, and a lock-in question for you.
Anthropic spent the back half of September quietly assembling something bigger than another feature drop. On September 23 it launched Claude Marketplace, a unified discovery hub inside Claude for connectors, agents, products, and — this is the part worth pausing on — consulting services. Two days later it opened a submission portal so any developer can list an MCP connector or a packaged plugin bundle, with automated validation and safety scanning attached. By September 27, the trade press had caught up and was calling it, accurately, Anthropic’s answer to a Google Play Store for AI.
If you’ve been reading this site’s tool reviews, you already know why this matters more than the average platform announcement: half the value of a tool like Claude Code or Claude in Cowork comes from what it can plug into. A marketplace is Anthropic betting that the way to win the next stage of this fight isn’t a better model, it’s better distribution for the stuff built on top of the model.
What actually launched
Strip away the “app store for AI” framing and here’s the concrete shape of it:
For users, Claude Marketplace is a browsable, searchable catalog inside Claude — at claude.com/marketplace — covering three tiers. Connectors and plugins (2,000-plus of them) come from companies you already deal with: Atlassian, Google, Microsoft, Notion, Salesforce, and others. A layer above that sits Claude-powered agents and products from partners like Cursor, CrowdStrike, Harvey, Lovable, and Snowflake — meaning some of the tools this site reviews are themselves now marketplace listings inside Claude, not just standalone products. At the top, a services layer lists systems integrators and consultancies (Accenture, Boston Consulting Group, Deloitte) offering implementation help, aimed squarely at enterprise buyers who want someone to hold their hand through a Claude rollout.
For developers, the September 25 piece is the submission portal: build a connector on MCP (Model Context Protocol) or package an Agent Skill, submit it, and it goes through automated validation and safety scanning before it’s listed. Anthropic also turned on version-discovery and install-metrics analytics for publishers, which is the unglamorous plumbing that makes a marketplace sustainable — developers who can see install numbers have a reason to keep investing in a listing instead of abandoning it after launch.
For enterprise buyers, the strategic hook is “committed spend”: Anthropic is letting large customers apply Anthropic contract dollars they’ve already committed to toward marketplace purchases from ecosystem partners. That’s a distribution deal wearing a UX feature’s clothes — it makes Claude the default checkout counter for a chunk of enterprise AI software spending, not just the model underneath it.
Why Anthropic is doing this now
Two pressures explain the timing. First, discovery was genuinely broken. Claude has had MCP connectors and Agent Skills for a while, but finding out what existed meant reading GitHub READMEs or trusting word of mouth — exactly the kind of fragmented mess a marketplace exists to fix. Second, and more pointed: OpenAI tried something similar with a GPT Store and it largely fizzled, criticized for weak discovery, thin incentives for builders, and a flood of low-effort listings. Multiple outlets covering this launch explicitly framed Anthropic’s version as designed to avoid that fate, by opening submissions to third-party developers from day one instead of restricting the marketplace to Anthropic’s own first-party integrations.
Whether that actually works depends on the unglamorous stuff: how good the safety scanning really is, whether install-metrics data is enough to keep third-party developers building, and whether “2,000-plus connectors” turns into 2,000-plus connectors anyone actually uses versus 2,000-plus abandoned listings a year from now. Early is early.
What this means if you’re not Anthropic
Three practical takeaways, in order of how soon they’ll touch your workflow:
If you already use Claude Code, Claude in Cowork, or Claude’s Slack integration, go look at the marketplace before you build another custom integration by hand. There’s a real chance the CRM sync, the analytics pull, or the ticketing hookup you were about to wire up yourself already exists as a vetted connector. That’s genuinely useful and it’s the reason this is a “go check this out” item rather than a “wait and see” one.
If you evaluate AI coding and business tools for a living — which, if you’re reading this site, you probably do — pay attention to which of your existing vendors show up as marketplace listings versus which stay outside it. Cursor and Lovable both appear as partner products inside Claude’s marketplace, which is a data point about how these companies see their own competitive position: partnering with the model layer rather than only competing against it. That’s worth factoring into how sticky you expect any of these tools to be over the next year.
And if you’re the one building something — a connector, a Skill, a workflow template — the submission portal is a real distribution channel now, not a side project. The bar to clear is Anthropic’s automated safety scanning plus whatever quality signal install-metrics data ends up rewarding. Early listings in a new marketplace category tend to get outsized visibility before the shelf gets crowded; if you have something worth listing, this is the moment for it, not six months from now.
The one thing to watch
A marketplace’s real test isn’t the launch-day partner list — Atlassian and Salesforce were always going to show up, they show up for every platform’s launch partner slide. The test is whether an independent developer with a genuinely useful MCP connector can get discovered without a business-development relationship with Anthropic. That’s the exact thing that sank comparable efforts before, and it’s the one number — not the 2,000-plus headline — worth checking back on in a few months.
Related guides
founder · 7 min read
Your AI Coding Tool Just Started Keeping Your Data Longer
GitHub Copilot chat retention goes from 28 days to the life of your account on Sep 28. And Claude Fable 5.1 now requires retention to run at all.
founder · 7 min read
Anthropic Files to Go Public: What an IPO Means for the Tools You Build On
Anthropic raised $65B at a $965B valuation and confidentially filed for an IPO. Here's what it actually changes for founders building with Claude-powered tools.
founder · 8 min read
Anthropic put three agents on one project. They sabotaged each other.
New Anthropic research on what happens when AI agents share a codebase. Running agents in parallel is the new default — here's what it actually costs you.
Enjoying this guide?
Get weekly practical guides, plus tool updates and implementation playbooks.