Microsoft recently released the Managed Extensibility Framework (MEF) which allows developers to add hooks into their applications so that the application can be extended at runtime.

Using MEF is a 2-step process: one step is performed by the application developer who adds attributes or code at defined points in the application.  At these points, the application searches for extensible objects and adds or call them to the application at runtime. 
The second step is by third-party developers who use the MEF application programming interface (API) to define classes in an “extension” assembly as extensible so that they will be discoverable by the above-mentioned applications.

The two steps are loosely-coupled, meaning neither the application nor the extension assembly needs to know anything about the other.  We don't even need to set a reference from one project to another in order to call across these boundaries.

I can think of two scenarios where this technology would be useful.

In scenario 1, an independent software vendor develops and sells a package with many pluggable modules.  Customers may choose to buy and install one module or all modules in the package.  For example, an Accounting package may offer General Ledger, Accounts Payable, Accounts Receivable Payroll and Reporting modules, but not all users will want to pay for every module.  By using MEF, the software could search a well-known directory for any module assemblies (flagged as extensions by MEF) and add to the menu only those that are installed.  With MEF in place, more modules could be added at a later time with no recompiling.

In scenario 2, developers create and deploy an application with a given set of functionality and create points at which other developers are allowed to extend the application using MEF.  By publishing these extendable points, they can allow developers to add functionality to the application without modifying or overriding the original source code.  This is a much safer way of extending functionality.  Extensions could be anything from new business rules or workflow to additional UI elements on forms.

All the extensions happen at runtime and MEF gives developers the ability to add metadata to better describe their extension classes.  By querying this metadata, we can conditionally load only those extensions that meet expected criteria.  The best part of this feature is that we can query a class's metadata without actually loading that class into memory.  This can be a huge resource saving over similar methods, such as Reflection.

MEF is currently released as Community Technology Preview (CPT), so the API is likely to change before its final release.  You can download the CTP and read more about it at http://code.msdn.microsoft.com/mef.   By learning it now, you can be prepared to add extensibility to your application when MEF is fully released.