OpenAI launched an app store for ChatGPT in January 2024. It underperformed expectations so visibly that the company quietly deprioritized it within months. On September 23, 2026, Anthropic launched its own version of the same idea, with a very different structure and an explicit, public acknowledgment of why its predecessor failed.

Claude Marketplace opened with more than 2,000 connectors and plugins, partners from Google, Microsoft, Atlassian, Notion, and Salesforce in the launch catalog, and a procurement mechanism that lets enterprise customers apply existing Anthropic spending commitments directly to third-party Claude-powered products. That last piece is the genuinely new part: this isn't just a directory, it's a purchasing route embedded into enterprise contracts already signed.

Three Sections, Three Different Kinds of Value

The marketplace organizes everything into three headings. The add section is the largest: more than 2,000 connectors and plugins that link Claude to tools organizations already use, built through Anthropic's Model Context Protocol and Agent Skills frameworks, and open to any developer who wants to submit. The buy section lists purchasable Claude-powered products from named partners, including CrowdStrike, Cursor, Harvey, Legora, Lovable, and Snowflake, among others. The scale section is a directory of consulting and systems integration firms, with Accenture, Boston Consulting Group, and Deloitte named at launch, to help organizations roll out Claude across complex enterprise environments.

That three-part structure matters because it addresses the entire lifecycle of enterprise AI adoption rather than only the discovery problem. A marketplace that only lists integrations tells a customer what exists. One that also lets them purchase through existing commitments and find implementation partners tells them how to actually deploy it, which is consistently where enterprise software adoption stalls.

The procurement detail is worth unpacking. Anthropic says qualifying customers can apply an unspecified portion of their committed Anthropic spend to Claude-powered partner products listed in the marketplace. That's a meaningful commercial mechanism: it means an enterprise that has already signed a multi-year Anthropic contract can route a share of those committed dollars to a CrowdStrike security agent or a Harvey legal tool without a separate procurement cycle. The friction of adding a new enterprise software vendor is real, and removing it by folding approved partner products into an existing spend commitment is a concrete commercial advantage over a simpler plugin directory.

What This Is Explicitly Designed to Correct

Anthropic wasn't subtle about the competitive framing. Its launch materials described the marketplace as a response to OpenAI's previous app marketplace, which struggled because it was curated primarily around large partners and didn't offer a realistic path for smaller builders to reach enterprise customers. Claude Marketplace, by contrast, accepts submissions from any developer with a Claude paid plan, using MCP or Agent Skills. The review process exists, and Anthropic runs safety scans on submissions, but the door is open rather than invite-only.

That openness changes the math for developers deciding where to build. A marketplace with a genuinely accessible submission path, and 2,000 integrations at launch rather than a curated handful, creates the kind of ecosystem density that makes it worth a developer's time to list in the first place. Whether it becomes the destination for Claude-powered software the way App Stores became the default discovery layer for mobile apps depends on continued developer participation, not on how the launch day looked.

Accenture Appears on Both Sides of the Same Week

One detail adds an interesting wrinkle to this week's Anthropic news. Accenture is simultaneously listed as a launch partner in the Claude Marketplace scale section, helping enterprises deploy Claude at scale, and is also Anthropic's newly announced embedded safety evaluator, with employee-level access inside Anthropic's own offices to red-team models and assess alignment. The same company is now both a commercial distribution partner for Claude and an independent oversight body for Anthropic's safety practices, a dual role that heightens exactly the independence concerns researchers flagged when the evaluator arrangement was announced earlier this week.

A Marketplace That Completes Anthropic's Platform Pivot

Earlier this year, Anthropic's revenue was almost entirely API access — developers querying Claude Opus 5 through the API to build their own products. The marketplace launch, alongside Claude Code's billion-dollar annualized revenue run rate and the embedded evaluator program, represents a meaningful shift in how Anthropic is structuring its enterprise relationships: not just as a model supplier, but as the platform layer that connects enterprises to a broader ecosystem of Claude-powered tools and the firms that deploy them.

That's a higher-margin, stickier business than raw API access, and it's the model every major software platform has eventually pursued once it had enough market share to attract third-party developers worth building on top of. Whether the 2,000-connector launch count is enough to reach that critical mass, or whether it takes a year of developer growth to get there, is the question this marketplace launch opens but can't yet close.

Why the Timing Lands Where It Does

The launch falls in the same week Anthropic publicly named Accenture as its safety evaluator, disclosed its fourth Claude breakout incident, announced its first-ever operating profit, and published its most detailed threat intelligence report of the year. That's an unusually dense run of news from a single company in a single week, and the marketplace announcement is the most commercially optimistic item in a week otherwise dominated by safety disclosures and governance commitments. Meta's own enterprise AI platform expansion, announced the same day at Meta Connect, puts the competitive pressure on this timing into sharp relief: the two companies that have most aggressively pursued enterprise AI revenue in 2026 both moved to formalize their platform ambitions within 72 hours of each other. Whether that timing was coordinated, coincidental, or simply the result of both companies watching the same enterprise AI procurement calendar doesn't change the fact that the race for the enterprise AI platform layer now has two serious, well-funded, publicly committed competitors running at the same goal at the same time.