Anthropic Opens Directory Submission Portal for Claude Plugins

Anthropic opened a directory submission portal on September 25, 2026, giving developers on paid Claude plans a way to submit plugins to the Claude directory, track them through review, and see usage analytics once they are live.

Anthropic describes plugins as packages of MCP connectors, Agent Skills, or both, and as the main way to build third-party extensions for Claude. Once a plugin is approved, it is listed in the Claude directory. The company said millions of people connect Claude to their apps, work tools, and data every day.

Two Submission Paths

The portal accepts two forms of submission. The first is a single MCP connector that points to a remote MCP server. The second is a plugin bundle that combines MCP servers and skills, hosted on GitHub and submitted as a repository. In Claude Code, Anthropic’s command-line coding tool, plugins can also include LSPs, commands, hooks, and agents.

Whichever path a developer chooses, the portal guides the plugin from submission to launch.

Validation, Review, and Post-Publish Analytics

Each submission is automatically validated and safety-scanned as soon as it is submitted, a step Anthropic says lets developers catch issues early. Developers can see where a plugin sits in the review process, the results of the safety scan, and recommended changes. Once a plugin is approved, the developer decides when to publish it in Claude.

After publication, usage analytics show installs by product surface and by version, while discovery analytics show how often a listing is viewed and which searches lead people to it. Anthropic says the figures are meant to help developers prioritize fixes and features and to refine their listings so they reach new users.

Anthropic’s build documentation states that the company checks each submission before it is listed and that there is no partner program to apply to first. The directory at claude.ai/directory is the catalog inside Claude where people on paid plans find and add plugins and connectors.

What a Plugin Packages and How It Is Built

The documentation describes a plugin as a package people add once and then use on claude.ai, in the desktop and mobile apps, in Cowork, and in Claude Code. A plugin packages whatever the integration needs, in any combination: an MCP connector, included when Claude needs to reach the developer’s product or data; skills, which teach Claude workflows; an MCP App, included when the connector should show interactive UI in the conversation; and commands and agents, which provide named actions or specialists Claude can delegate to. A plugin containing only one of those pieces is complete.

The documented build sequence runs in order: lay out the plugin folder and manifest, write the skills in SKILL.md files, implement and test the MCP server, optionally add interactive UI, test the plugin on each surface, publish to the directory, and measure usage.

Most of the tooling for building and checking a plugin lives in Claude Code. The claude plugin validate command checks the manifest and component files, claude --plugin-dir loads a plugin from a local folder for one session, and claude plugin eval runs developer-written eval cases comparing Claude’s results with and without the plugin, with each run counting as a real model call against plan usage or an API bill. Inside a session, /skill-doctor reports what each installed skill costs in context and how often Claude has invoked it. Anthropic also publishes a skill-creator skill, which drafts a skill with the developer and runs it on test prompts, and the plugin-dev and mcp-server-dev plugins, whose skills walk Claude through building a plugin or an MCP server.

MCP 2.0 Support and Extensions

Anthropic said Claude supports the latest MCP specification, commonly referred to as MCP 2.0, which includes a stateless core. The specification version dated 2026-07-28 defines the Model Context Protocol as an open protocol for integration between LLM applications and external data sources and tools. It uses JSON-RPC 2.0 messages between hosts, clients, and servers: hosts are LLM applications that initiate connections, clients are connectors within the host application, and servers are services that provide context and capabilities. Its base protocol specifies stateless, self-contained requests with per-request capability negotiation.

Beyond the core protocol, MCP defines optional extensions that add modular, specialized, or experimental functionality; extensions are opt-in and require explicit support from both client and server, negotiated during initialization. Anthropic highlights two extensions for plugin developers: MCP Apps, for interactive UI inside chat, and Enterprise Managed Auth, for zero-touch OAuth for enterprise users. The MCP project describes MCP Apps as letting servers display interactive elements such as charts, forms, and video players inline within conversations, and Enterprise-Managed Authorization as a framework for enterprise environments that require centralized access control. Anthropic said support for more MCP features and extensions is coming soon.

Rollout and Existing Listings

Over the coming weeks, one discovery experience will roll out across Claude and Claude Code, Anthropic said. Skills and MCP connectors remain as building blocks, and the Claude directory will continue to list them. In the future, developers will be able to turn an existing connector listing into a plugin. Anyone with an existing skill, connector, or plugin on the Claude directory does not need to make any changes.