Claude Now Has a Marketplace. What Claude Plugins Mean for Your Business (and What to Check Before You Install One)

On 23 September 2026, Anthropic opened the Claude Marketplace: one place inside Claude to find plugins, connectors, Claude-powered products and service partners. More than 2,000 connectors and plugins were listed on day one, from names like Atlassian, Google, Microsoft, Notion and Salesforce.

For a small business team, that sounds like a gift. Claude can now plug straight into the tools you already run. It also means your team is now a few clicks away from handing an AI assistant access to your inbox, your files and your customer records. This guide covers what Claude plugins for business actually are, which ones are worth switching on first, and the security checks to run before anyone in your team clicks "install".

Quick answer: Claude plugins for business are add-ons that give Claude role-specific skills, commands and connections to your other software, such as your email, documents or CRM. The Claude Marketplace, launched 23 September 2026, lists over 2,000 of them. Install only what solves a real task, grant the least access possible, and have one person approve every install.

What are Claude plugins, and what is the Claude Marketplace?

Plugins are not brand new. Anthropic introduced plugins for Claude Cowork, its desktop assistant for knowledge work, in late January 2026, then expanded them for business teams on 24 February 2026. What is new is the marketplace that puts them all in one catalogue.

Here is the vocabulary in plain English:

Term	What it is	Business example
Connector	A link that lets Claude read from, or act in, another piece of software	Claude reads your shared drive or creates a task in your project tool
Skill	A packaged set of instructions that teaches Claude how your team does a specific job	"Write a quote in our format", "Summarise a tenancy application our way"
Plugin	A bundle of skills, slash commands and connectors set up for a role or department	A finance plugin, a sales plugin, an HR plugin
Marketplace	The catalogue where you find and add all of the above, plus Claude-powered products and consultants	Browse, add, and manage what your team uses

According to Anthropic's announcement, the marketplace has three sections: connectors and plugins, agents and products from companies such as CrowdStrike, Cursor, Harvey and Snowflake, and service partners such as Accenture, Boston Consulting Group and Deloitte. Larger customers with a committed Anthropic spend can put part of that commitment toward listed products.

The headline for a small business is simpler: Claude can now work inside your other systems, not just in a chat window. That is where the real time savings live, and it is also where the real risk lives.

Why this matters for a small business team

Most small teams use AI the same way: someone copies information out of one system, pastes it into a chat, then copies the answer back. It works, but it is slow, and it is exactly how customer data ends up in places it should not be (we covered that in what not to paste into ChatGPT).

Plugins and connectors remove the copy and paste. Claude reads the source directly, does the work, and can often write the result back. In practice that means:

• Less swivel-chair admin. Pulling figures from a spreadsheet into a report, or notes from a meeting into a task list, without re-typing.
• Consistency. A skill makes Claude follow your format, your tone and your checklist every time, rather than whatever each staff member happens to type in.
• Speed to value. Many connectors are already built for the software you pay for, so there is less custom build.

The trade-off is access. A connector that can read your email can read all of your email. A plugin that can update your CRM can also update it wrongly. The question is not "should we use plugins?" It is "which ones, with what access, and who decides?"

Which Claude plugins should a small business switch on first?

Start with the boring, high-volume work. The best first plugins connect Claude to systems your team already lives in and take on tasks that are repetitive, low risk and easy to check.

Priority	Type of connection	Good first tasks	Risk level
1	Documents and shared drive	Summarise contracts, find policies, draft from templates	Low to medium (read-only where possible)
2	Project or task management	Turn meeting notes into tasks, weekly status summaries	Low
3	Calendar and meeting tools	Prep notes before meetings, action lists after	Medium (calendar details are personal information)
4	Email	Draft replies, triage the inbox, find old threads	Medium to high (start with drafting only, never auto-send)
5	CRM and finance systems	Pipeline summaries, follow-up lists, report prep	High (customer and financial data, needs a policy first)

Two rules make this table work:

1. Read before write. Start every connector with read-only access if the software allows it. Let Claude prove it gets the answers right before it is allowed to change anything.
2. Draft before send. Anything that leaves your business, such as an email, a quote or a message to a customer, should be drafted by AI and sent by a human. That one rule prevents most of the embarrassing mistakes.

The security checklist: what to check before you install a Claude plugin

I have spent 25 years in IT and cyber security, and this is the part of the announcement that made me sit up. Anthropic's marketplace post explains how builders can submit connectors and plugins, but it does not describe how third-party listings are security reviewed. That does not mean they are unsafe. It means the checking is on you.

Treat a plugin the way you would treat any new app that asks for access to your business systems. Before anyone installs one, run these checks:

1. Who built it? Is it published by the software vendor itself (for example, the company that makes your CRM) or by an unknown third party? Vendor-built connectors carry less risk.
2. What access does it ask for? Read-only or read-and-write? One folder or the whole drive? One mailbox or every mailbox? Grant the smallest scope that does the job.
3. Where does the data go? Does the connection send your information to anyone other than Anthropic and the software you are connecting? If the listing cannot answer that, do not install it.
4. Can it act without a human? Anything that can send, delete, pay or publish on its own needs a human sign-off step, or it should not be connected yet.
5. Who approved it? One named person in your business should approve every plugin and connector. If staff can add their own, you have recreated shadow AI with more access.
6. Can you see what it did? Use admin controls and activity logs where your plan provides them, so you can answer "what did the AI touch?" if something goes wrong.
7. Can you switch it off fast? Know how to disconnect a connector and revoke its access in the source system, not just in Claude.

If this sounds like overkill, consider what an AI agent with too much access can do. We unpacked a real example in what the Medicare AI agent breach means for your business. The lesson there applies here: least privilege, logging and human sign-off are not optional extras.

Admin controls: who should be allowed to install plugins?

When Anthropic expanded plugins for business teams in February 2026, it added admin features aimed at exactly this problem. According to Anthropic, admins can set up private, company-specific plugin catalogues, control which teams can access which plugins, provision plugins for users, and manage connector and MCP controls. Some of these features depend on your Claude plan, so check what your plan includes before you rely on them.

For a small business, the practical setup looks like this:

Decision	Recommended approach
Who can install plugins?	One admin (the owner or a trusted operations lead), not everyone
How are plugins shared?	A short approved list, shared with the teams that need them
Who can connect email or CRM?	Only after an AI policy is in place and the access scope is agreed
How often do you review?	Quarterly: remove anything unused, check access is still minimal

If you do not yet have rules for how your team uses AI, write those first. Our guide on how to write an AI policy walks through a simple version you can put in place in an afternoon. Australian businesses should also keep the Privacy Act in view. From 10 December 2026, organisations covered by the Australian Privacy Principles must explain in their privacy policy when they use computers to make decisions that significantly affect people. Connecting AI to customer data makes that question real.

A simple 30-day rollout plan for Claude plugins

You do not need a big project. You need a short, controlled trial.

Week 1: Pick the problem. Choose two or three tasks your team repeats every week that involve documents or task management. Write down roughly how long they take today.

Week 2: Connect one system, read-only. Add a single connector, such as your shared drive, with the narrowest access possible. Build or install one skill that does one task your way.

Week 3: Measure and check. Compare the time taken and the quality of the output against week 1. Have a human check every result. Note anything Claude got wrong.

Week 4: Decide. Keep, expand or remove. If it earned its place, add the next connection on the priority table. If not, switch it off and try a different task.

This is the same "start small, prove it, then scale" approach we recommend in our 5-step AI adoption plan. The marketplace makes it easier to add tools. It does not change the order you should add them in.

FAQ

What is the Claude Marketplace?

The Claude Marketplace is Anthropic's catalogue for adding connectors, plugins, Claude-powered products and service partners to Claude. It launched on 23 September 2026 with more than 2,000 connectors and plugins from companies including Atlassian, Google, Microsoft, Notion and Salesforce.

Are Claude plugins safe for business?

They can be, if you control them. Check who built the plugin, grant the least access possible, start read-only, keep a human in the loop for anything that sends or changes data, and have one person approve every install. Anthropic's marketplace announcement does not describe a security review process for third-party listings, so do your own checks.

What is the difference between a Claude plugin and a connector?

A connector links Claude to another piece of software so it can read from or act in it. A plugin is a bigger bundle that can include connectors, skills and slash commands set up for a particular role or department, such as finance or sales.

Do I need technical skills to use Claude plugins?

Not for the basics. Adding a vendor-built connector or plugin is designed for business users. Building custom skills, setting access scopes properly and connecting systems that hold customer data is where technical and security experience pays off.

Should I connect Claude to my email and CRM?

Eventually, perhaps, but not first. Email and CRM systems hold your most sensitive customer information. Start with documents and task management, put an AI policy in place, and only connect email or CRM with tight access and a draft-not-send rule.

Does using Claude plugins affect my Privacy Act obligations?

It can. If a connected AI tool handles personal information, your existing privacy obligations still apply. From 10 December 2026, organisations covered by the Australian Privacy Principles must also disclose in their privacy policy when automated decisions significantly affect individuals.

Your next step

The Claude Marketplace makes it easy to give AI access to your business systems. That is the opportunity and the risk in one sentence. The businesses that win with Claude plugins for business will be the ones that pick the right tasks, grant the least access, and keep a human on anything that leaves the building.

If you want help choosing which connections are worth it for your business, and setting them up safely, book your free AI Game Plan Session. In one hour we map where AI will save your team the most time and what guardrails need to be in place first. It is the starting point for Your AI Growth Department, and how most of our AI consulting work begins.