Claude Plugins: Package, Submit, Go Live

Claude just opened a portal for plugins. This is the whole thing in two minutes, one drawing at a time.

Three terms first. MCP, the Model Context Protocol: an open standard that plugs Claude into outside tools and data. A skill: written instructions Claude loads to do one kind of task well. A plugin: a package that carries MCP connectors, skills, or both — picture it as a sealed box, and either or both can go inside.

There are two ways to submit one. Way one: a single MCP connector, pointed at your own remote MCP server — that's the whole submission. Way two: a plugin bundle, where your MCP servers and skills sit together in a GitHub repository, and you submit the repo.

You submit right inside Claude, from the directory's manage page, if you're a developer on a paid plan. The moment you hit submit, the plugin is checked and safety-scanned, so problems surface early. From there you can track it through review — submitted, in review, live — including what the safety scan found and any changes it recommends.

Once it's live, you can see it being used: installs by product and version, how often your listing gets viewed, and which searches lead people to it. Why now? Claude's developer team reports MCP usage across Claude products is up 110x this year, and plugins are becoming the way to build for Claude.

Your Turn: Paste into Claude: "I use [name a tool] every week. Help me plan a Claude plugin for it: one MCP connector that exposes only the two or three actions I actually need, and one skill that says when and how to use them well. List what the connector needs access to, and flag anything it should not expose. Don't write code yet." Then check two things yourself: does every action trace to something you actually do with that tool, and does the connector reach anything the job doesn't need? If it does, cut it before you submit.

Sources: Anthropic, "Build plugins for Claude" (Sept 25, 2026), claude.com/blog/build-plugins-for-claude. The 110x MCP usage figure is from Claude's developer team on X (@ClaudeDevs), not the blog post.

0:00 The question, and three terms 0:23 What goes inside a plugin 0:40 Two ways to submit 0:57 Submit, scan, and track the review 1:17 Usage analytics, and why now 1:35 Your turn

Part of the Claude & Agentic AI series. @NikBearBrown

Deliberately Not Claimed

This video does not claim to cover every requirement or edge case in Anthropic's plugin submission process — it summarizes the core workflow as described in the cited blog post and developer post. Details such as review timelines, eligibility for the paid-plan requirement, or safety-scan criteria may change; check the linked source for the current, authoritative version.

SEO Keyword Tags

Claude plugins, Model Context Protocol, MCP connector, Claude skills, Claude plugin directory, how to submit a Claude plugin, Claude developer portal, Anthropic plugins, MCP server GitHub repo, Claude plugin review process, Claude plugin analytics, agentic AI tools, Claude MCP tutorial, building for Claude, Claude plugin safety scan

What this video covers

• 0:00 — The question, and three terms 0:23 What goes inside a plugin 0:40 Two ways to submit 0:57 Submit, scan, and track the review 1:17 Usage analytics, and why now 1:35 Your turn

Transcript

A lightly formatted transcript of the video, with timestamps so you can jump to any section on YouTube.

The question, and three terms 0:23 What goes inside a plugin 0:40 Two ways to submit 0:57 Submit, scan, and track the review 1:17 Usage analytics, and why now 1:35 Your turn

[0:00] Hello, this is Liam in for Bear. Claude just opened a portal for plugins. So, the question isn't only how you build one, it's how you get one live. Three terms: MCP, the Model Context Protocol, an open standard that plugs Claude into outside tools and data. A skill, written instructions Claude loads to do one kind of task well. And a plugin, a package that carries MCP connectors, skills, or both. Start with the plugin itself. Picture it as a sealed box. What goes inside? A plugin is a package. Inside go MCP connectors, which plug Claude into outside tools and data, and skills, which are written instructions Claude can follow. A plugin can carry either one or both. There are two ways to submit one. Way one is a single MCP connector.

You point it at your own remote MCP server, and that's the submission. Way two is a plugin bundle. You put your MCP servers and your skills together in a GitHub repository, and you submit the repo. You submit right inside Claude from the directories manage page, if you're a developer on a paid plan. The moment you hit submit, the plugin is checked and safety scanned, so problems show up early. Then you can track it. The portal shows where your plugin is in review, what the safety scan found, and any changes it recommends. Submitted, in review, live. Once it's live, you can see it being used. Installs by product and version, how often your listing gets viewed, and which searches lead people to it. Why does this matter now? Claude's developer

Team says MCP usage across Claude products is up 110 times this year, and that plugins are becoming the way to build for Claude. Your turn. Paste this into Claude. I use name a tool every week. Help me plan a Claude plugin for it. One MCP connector that exposes only the two or three actions I actually need, and one skill that says when and how to use them well. List what the connector needs access to, and flag anything it should not expose. Don't write code yet. Then check two things yourself. Does every action trace to something you actually do with that tool? And does the connector reach anything the job doesn't need? If it does, cut it before you submit. Cloud plugins package submit go live at Nick Bear Brown.