Anthropic Turns Claude Into a 2,000-Tool Marketplace
Anthropic launched the Claude Marketplace on September 27, a single directory bringing together more than 2,000 plugins and connectors from partners including Atlassian, Google, Microsoft, Notion and Salesforce, Claude-powered agents and products from CrowdStrike, Cursor, Harvey, Lovable and Snowflake, and consulting services from Accenture, Boston Consulting Group and Deloitte.
The notable design choice: it’s open. Any developer or company can apply to list, with no partner-program gate required — a structure Anthropic is explicitly framing as a lesson learned from OpenAI’s app store, which struggled to gain traction. In Anthropic’s words, the goal is to make it “easier for teams to discover products that already work with Claude, while giving developers and partners a direct way to reach existing Claude customers.”
The On-Ramp for That Marketplace Opened Two Days Earlier
On September 25, Anthropic opened a self-serve directory submission portal so any developer on a paid plan can submit a plugin — either a single MCP connector pointing at a remote server, or a bundle combining MCP servers and Agent Skills hosted on GitHub — and track it through automated validation and safety scanning.
Once approved, developers control when their listing goes live and get analytics afterward: installs by product surface and version, plus the search terms actually driving people to find it. No partner-program application required, and MCP 2.0 is supported from day one.
Claude Tag Can Now Use Your Own Connectors in a Channel
Claude Tag in Slack now lets individual users bring their personal connectors — Google Drive, a calendar, a CRM login — into a shared channel alongside whatever the admin has already provisioned for the whole team. Each person decides whether Claude screens its answers before posting or sends them automatically, and enterprise admins can force review for everyone.
Anthropic’s framing: “You decide what Claude reaches and what the channel sees, the same as when you use your connectors in a direct message.” It’s rolling out on Team plans first, with Enterprise access to follow.
In the DRC, Claude Turned a Day of Paperwork Into Under an Hour
WHO AFRO staff responding to the Bundibugyo Ebola outbreak in eastern DRC built a Claude workflow that pulls case and lab figures out of district PowerPoint decks, compares them against the prior day, flags trend changes, and drafts the situation report — cutting daily prep from a full day to under an hour. Teams are also now running several outbreak forecasts at once, where time constraints previously allowed only one, and using Claude Science to assemble and compare viral genome sequences from plain-language prompts.
The honest caveat, echoed in the coverage: humans still decide which blood samples get tested and which numbers make the report. The clearest win so far is faster paperwork, not automatically better outbreak decisions — a distinction worth keeping as these stories get cited as evidence of AI’s real-world value.
A Platform Launch and a Proof Point, in the Same Week
Three of today’s stories are one story wearing different hats. The submission portal, the marketplace it feeds, and personal connectors landing in Slack are all the same bet: make Claude the layer everything else plugs into, and make it as easy as possible for outside developers to do the plugging. Opening it wide — no partner-program gate, automated review instead of a hand-curated shortlist — is a direct answer to OpenAI’s stalled app store, and it’s a bet that openness scales safely at 2,000+ listings and counting.
The DRC story is the unglamorous version of the same thesis: not a flashy model benchmark, but a WHO team going from a full day of report-writing to under an hour during an active outbreak. That’s the kind of proof point a platform strategy needs to be more than plumbing. Worth watching as the marketplace scales: whether automated safety scanning holds up review quality once thousands of listings are the norm rather than the pitch.