The Gearset MCP server lets your AI coding assistant talk directly to Gearset. If you are already building changes with tools like Claude Code or Cursor, you can now start the process of creating new pipeline features without leaving your development environment.
Getting started with the Gearset MCP
To get set up and connected with the Gearset MCP server, read our documentation here. Instructions on how to manage access to individual MCP tools via role-based access control (RBAC) can be read here.
Prerequisites
Before your AI assistant can automatically open a new pipeline feature, you'll need a Gearset Pipeline set up with at least one development environment created and linked to a static environment.
You'll also need to make sure that your Gearset Account has the permissions and entitlements needed to create new pipeline features within that pipeline.
How does it work?
You've tested your changes in your developer environment and are ready to get a new feature created and moving through your pipeline - the Gearset MCP can help you move those changes into a feature branch. Simply ask your AI assistant to create the feature and Gearset will create the feature branch and start a comparison you can jump straight into - with the items you've been working on already selected.
Trigger your AI assistant to create a new feature with a prompt like the below, filling in which changes you'd like to include in the feature.
Create a pipelines feature called "my-feature" with these changes...
Gearset will then create the feature and launch a comparison. Once this is finished you'll be provided with a URL to open it so that you can view the results and change item selections.
Once you've verified the selected changes and committed to the newly created feature branch, you can return to your agent to raise a PR for the feature or continue working in the Gearset UI.
Precision selection is not yet supported and you may need to add more items to the comparison if dependencies aren't appearing, so be sure to double check the selected items.
We are continuing to improve this experience so please share any feedback in our in-app chat.
Limitations
GitFlow pipelines are not currently supported.
Attaching a ticket to the generated feature is not currently supported through the Gearset MCP server.
Item pre-selection only supports metadata items (i.e. not CPQ, Revenue Cloud, or Vlocity items), and only selects at an entire-object level.
Tips and tricks
This feature works best when your AI assistant has solid knowledge of which items need to be included in this pipelines feature; perhaps from a ticket your assistant has access to, or from the same conversation in which your assistant made the original changes in your developer environment.
Comparisons launched via the Gearset MCP server don't perform any automatic dependency discovery, so you'll need to ensure your AI assistant is aware of what associated items also need comparing (such as Profiles and Permission Sets), or add these to the comparison yourself in the UI.
Specifying items to include in the feature will compare the item and initially select it for deployment in the resulting comparison. This will currently select the entire item; if specific changes within the item are required (for instance if there are also other changes to the item not required for this feature), you'll need to use the comparison UI to unselect the entire item and reselect just those sections.