Transform Claude: Build Your First Plugin

In one line, Claude Plugins package all of Claude’s existing tools into one “role.”

Using Plugins is essentially a way to create a mini Claude employee with your existing tools.

For example, if you have 3 MCPs connected to Claude and 2 custom Claude Skills, you can combine them into a single Plugin that knows your workflows.

What most people don’t know is that Anthropic already has 24+ Plugins installed in your Claude app.

These pre-built Plugins contain custom slash commands, agents, MCP servers, and hooks in one place across different roles.

A practical example of a Plugin prompt for marketing:

“Hey Claude, use my marketing Plugin to create a custom Canva infographic with my brand.”

Claude then uses your Canva MCP connection and brand skills to finish the task end-to-end.

The cool thing is, you can also build custom Plugins and don’t have to rely only on the ones Anthropic has built.

Now that you have a base understanding of what Plugins actually are, let’s build your first custom Claude Plugin:

To build an effective Claude Plugin, you need to start with a proper planning phase.

Think of this like writing a job description - the clearer you are about what this Plugin does, who it is, and what skills it has, the better it will perform.

Answer these three questions:

What is the role? Give your Plugin a clear job title. (marketing assistant, personal assistant, etc.)

What MCPs and connectors does it need? List the external tools it should have access to. (Gmail, Notion, Canva, Slack, etc.)

What skills does it need? List any existing Claude Skills that should be baked into this Plugin. (brand voice, writing style, formatting, etc.)

With your plan locked in, open Claude Cowork on desktop and let it do the heavy lifting.

You don't need to write a single line of code. Just describe the Plugin you want to build, and Cowork handles the rest.

Here's the prompt to get started:

Once Cowork builds it, you'll see the full plugin structure - skills, connectors, slash commands, and hooks all bundled together in one place.

Pro Tip

If you don’t know what Plugin to initially build, you have two options:

1) Prompt Cowork: “Based on everything you know about me, what Plugin would be most helpful in my daily workflow?”

2) Online: You can surf the web for existing Claude Plugins. If you find one you like, don’t download it blindly. Instead, copy/paste the description into Cowork and say:

“I found this Plugin online, help me reverse-engineer how it was built so we can build one together.”

Your Plugin is now fully built. Next, let's make sure it actually works.

Installing your plugin:

Go to Claude desktop → Cowork → Customize → Plugins and activate your new Plugin.

If you built it directly in Cowork, it should already be there.

Testing it:

Start simple. Run a prompt that covers the plugin's core function and see how it responds. Don't try to stress test everything at once - pick the one thing your plugin is supposed to do best and verify it does it well first.

A good first test prompt:

If something isn't right:

Don't rebuild from scratch. Go back into Cowork and tell it exactly what's off.

Treat it like iteration - one fix at a time until the output matches what you planned.

Once it's working:

Set up a custom slash command library so you can trigger the full plugin with a single word.

Type /[plugin name] and Claude activates everything - connectors, skills, context.

Marketing Plugin example

To start using your new Plugin, just start calling on it directly in chats.

Example prompts

“Hey Claude, use my marketing Plugin for an SEO audit.”

“Use my marketing Plugin to write sales copy for [x].”

It was a huge week in AI, so you won’t want to miss this!

Anthropic closed a $65B Series H on May 28 at a $965B post-money valuation, surpassing OpenAI's $852B mark.

Anthropic has also started the official IPO filing process.

Anthropic’s latest model upgrade is here.

Opus 4.8 brings upgrades to coding, 4x fewer hallucinations, a new “Ultracode” mode, and dynamic workflows that spin up dozens of AI agents.

You read that right.

Anthropic is aiming to release a public version of their Claude Mythos model - the world’s strongest cybersecurity LLM to date.

They’ve announced a massive expansion to Project Glasswing - the project dedicated to developing Claude Mythos.

The world’s most powerful AI agent just got a massive upgrade.

Hermes agent is now available as a desktop app - no complex terminal setup required.

On June 2, President Trump signed an executive order requiring AI companies to provide the U.S. government early access (up to 30 days) to their most powerful frontier models before public release.

Composer 2.5 - xAI’s most powerful lightweight LLM is live in Grok Build now.

Jensen Huang just revealed the world’s most powerful laptop ever built.

Full introduction video:

GPT-5.6 is expected within the next week.

Keep your eyes peeled for any announcements regarding what will presumably be the world’s most powerful model.

As promised, we have a special announcement.

Exclusive for all newsletter members, we are launching a premium AI asset library.

This asset library will contain custom AI workflows, agents, prompts, skills & much more - built to help you level up with AI (100% free for our newsletter community).

Public access and official links will be sent in the coming publications - we’re super excited to share what we’ve been building.

Stay tuned!