Microsoft’s Copilot Plugin Catalog: What IT and AI Leaders Should Do Before Plugins Scale

Glowing glass plugin modules connected to a central hub carrying the Microsoft Copilot logo, with the Stellium and Microsoft Solutions Partner logos and an Insights label
Table of Contents

Last reviewed 17 November 2026. This article reflects Microsoft’s announcements ahead of and during Microsoft Ignite 2026. We’ll update it as Microsoft publishes further details.

Most organisations that adopted Microsoft Copilot early now face a familiar problem. Agents were built by one team, connectors were enabled by another, prompts and instructions were shared in documents and chats, and employees were left to work out what was available and what they were allowed to use.

Microsoft’s answer is the plugin. Ahead of Ignite 2026, the company confirmed that skills, tools and agents will be packaged and governed together, and that a new public Plugin Catalog launches on 17 November 2026 to help organisations explore the wider ecosystem.

This is a useful step. It gives IT a single model for managing Copilot extensions and gives employees a clearer way to find them. But a catalog is a starting point for discovery, not a strategy. The organisations that benefit most will be the ones that decide early which plugins solve real business problems, and how those plugins will be approved, maintained and retired.

Key takeaways

  • A plugin is now the unit of extension in Copilot. It can combine skills, tools (including connectors) and agents in one package.
  • The public Plugin Catalog launches on 17 November 2026. It helps you explore Microsoft and wider ecosystem plugins before deciding what to enable.
  • The plugin registry is where governance happens. Admins approve, assign, block and manage plugins centrally in the Microsoft 365 admin center.
  • Navigation is changing for users. By late November 2026, the Agents, Agents & Skills and Customize entry points are replaced by a single Plugins option.
  • No technical migration is required, but preparation is. Update training, internal documentation and your approval process now.

What Microsoft has announced

Microsoft describes a plugin as a package that brings together the capabilities Copilot needs to complete a piece of work. According to the Microsoft Copilot blog, a plugin can include any combination of three components:

  • Skills: reusable instructions that help Copilot perform a specific task consistently.
  • Tools: the ability to retrieve information or take action in business systems, within defined inputs, outputs and permissions. Connectors sit within this category.
  • Agents: capabilities that reason toward a goal and orchestrate skills and tools to complete a workflow.

Three related changes sit around this definition.

1. The plugin registry. Introduced in September, the plugin registry is a single place to discover and govern plugins across Copilot experiences. Through the Agent 365 experience in the Microsoft 365 admin center, under Agents > Tools, admins can review Microsoft, custom-built and partner plugins, enable or disable them, and block or assign them to specific users and groups.

2. The public Plugin Catalog. Launching on 17 November 2026 on microsoft.com, the catalog lets organisations explore the broader plugin ecosystem before anything is enabled in their tenant. Think of it as a shop window. The registry remains the control point.

3. A new navigation for users. Plugins now appear directly in Copilot’s navigation, alongside the new Copilot Home that brings Chat and Cowork together. Microsoft’s message centre notice MC1484008 confirms that by late November 2026, users will select Plugins instead of Agents, Agents & Skills or Customize.

Microsoft also says the registry will extend to experiences such as Copilot Studio, GitHub Copilot and Microsoft Foundry in the future, and that Work IQ Developer Tools will support the full lifecycle of custom plugins, from building and validation to publishing through the registry.

Why this matters now

Until now, many organisations managed Copilot extensions as separate items. An agent was approved in one place, a connector in another, and reusable prompts lived wherever someone happened to save them. That approach works for a pilot. It does not work when dozens of teams start building.

The plugin model changes three things.

Discovery becomes easier, so demand will grow. When employees can browse available capabilities in one place, they will ask for more of them. IT teams should expect a rise in requests once the new Plugins entry point appears.

Governance becomes more consistent. One registry and one approval model reduce the risk of agents, connectors and instructions being managed under different rules. This is a real improvement for regulated organisations that need to show who approved what, and why.

Reusable knowledge becomes an asset. Skills are easy to overlook, but they may be the most valuable part of the model. A well-written skill captures how your organisation does a task, such as reviewing a contract clause, preparing a client briefing or summarising a project status. Once packaged as a plugin, that know-how can be reused and improved instead of rewritten by every team.

Discovery is the easy part

The catalog will make it simple to find plugins. The harder questions sit with the organisation.

Consider a hypothetical professional services firm with 2,000 employees. Its marketing team wants a CRM plugin to prepare client briefings. Its finance team wants a plugin that connects to the ERP system for month-end commentary. Its delivery teams want a set of skills for proposal writing.

Each request is reasonable. Together, they raise questions the catalog does not answer:

  • Which business systems should Copilot be allowed to read from or act on?
  • Who owns each plugin after it is approved, and who updates it when the underlying system changes?
  • Which groups should receive each plugin, and which should not?
  • How will the firm know whether a plugin is improving the work it was meant to support?

Without clear answers, the risk is not a lack of capability. It is a growing list of approved plugins that nobody maintains, and that employees stop trusting.

What IT and AI leaders should do next

1. Set an approval baseline before requests arrive

Define what a plugin must show before it is enabled. A short, consistent checklist is enough to start:

Approval areaQuestion to answer
Business purposeWhich workflow does this plugin support, and what outcome should improve?
Data accessWhich systems and data does it read from or act on?
PermissionsDoes it respect existing access rights, or does it need new ones?
OwnershipWho is accountable for the plugin, and who maintains it?
AudienceWhich users or groups should receive it first?
ReviewWhen will it be reassessed, and under what conditions would it be withdrawn?

2. Map your current agents, connectors and skills

Before the navigation changes, list what already exists in your tenant. Identify duplicates, unused agents and connectors with unclear owners. This inventory becomes the starting point for your registry and helps you avoid carrying old clutter into the new model.

3. Prepare users for the navigation change

Microsoft states that no setup or technical migration is required for the new Plugins entry point. Users will still notice the change. Update internal guides, training materials and adoption content that refer to Agents, Agents & Skills or Customize, and tell people what they will see and where to request new capabilities.

4. Start with skills for high-value, repeatable tasks

Pick two or three workflows where teams repeat the same work with inconsistent results. Capture the best version of that work as a skill, test it with a small group and refine it before wider release. Skills are low risk to pilot because they shape how Copilot works without necessarily opening access to new systems.

5. Use the catalog to shortlist, not to enable

Treat the public catalog as research. Shortlist plugins that match your priority workflows, then run them through your approval baseline and a controlled pilot before assigning them more widely.

6. Measure completed work, not plugin usage

The number of plugins enabled is not a measure of success. Track whether approved plugins reduce time on a task, improve output quality or reduce rework. Retire plugins that do not deliver.

A practical rollout plan

Phase 1: Prepare (now). Build your inventory, agree the approval baseline and update user communications.

Phase 2: Pilot (next 4 to 8 weeks). Enable a small set of plugins for defined groups, focusing on two or three priority workflows. Record results against the outcomes you agreed upfront.

Phase 3: Expand (based on evidence). Widen access to plugins that show measurable value. Assign owners, set review dates and publish a simple internal list of what is approved and why.

Phase 4: Operate. Review the registry regularly. Update skills as processes change, remove plugins that are no longer used and keep the approved list current.

Frequently asked questions

Do we need to take technical action for the new Plugins entry point?
Microsoft says no setup or technical migration is required. It recommends updating internal documentation, training and adoption materials that reference the old entry points.

Is the public Plugin Catalog the same as the plugin registry?
No. The catalog on microsoft.com helps you explore what is available. The registry, managed through the Microsoft 365 admin center, is where your organisation approves, assigns and governs plugins.

Can we build our own plugins?
Yes. Microsoft positions Work IQ Developer Tools as the route to build, validate, package and publish custom plugins through the registry. Microsoft’s admin guidance on managing plugins, skills and MCP servers is a useful starting point for administrators.

Where can we learn more at Ignite?
Microsoft lists sessions on plugin registry governance (BRK221) and the full plugin lifecycle (BRK237), alongside sessions on skills and Copilot connectors. Details are on the Microsoft Copilot blog.

Turn plugins into a governed business capability

The Plugin Catalog will make it easier for everyone to see what Copilot can do. That visibility is valuable, but it also raises expectations. The organisations that get the most from plugins will treat them as business capabilities with clear owners, clear purposes and clear measures of success.

Start with an approval baseline, a clean inventory and a few workflows that matter. Then let the evidence decide what scales.

If you’re planning how plugins fit into your Copilot roadmap, Stellium can help you define the governance model, prioritise the right workflows and run a controlled pilot. You can also read our view on Claude in Microsoft Copilot’s EU Data Boundary for another example of turning a Copilot announcement into a practical rollout plan.

Stellium

17 November 2026