Let’s say I want to fully automate a basic software development workflow using agent frameworks like Flowise, Langchain, CrewAI, or others. I’d use an MCP or a REST API to retrieve ticket details, fetch the relevant code from the repository, and ultimately generate a merge request.
However, before submitting the merge request, I need to modify the code. At that point, I have two options:
I can build my own flow using an MCP capable of reading and editing files, orchestrating multiple LLM calls to determine what needs to be changed and where in the code.
Or, I could leverage a tool like Cursor, which already does all of this, and simply expose its features through an MCP API.
The second option is far more efficient. Cursor already handles many of the complexities I’d otherwise have to reimplement myself — things like rule enforcement, code formatting, linting, and intelligent code editing. Exposing Cursor’s capabilities via an MCP would likely produce much better results with less effort.
+1 as time goes, a Cursor MCP will be increasingly needed, or else some customers needing this feature in their agentic workflow will migrate, if Cursor could possibly be also a standalone MCP server, a lot of possibilities opens for achieving a great number of customers, example: I create a team in Cursor, give each of my customers a key for Cursor and my software, then they need to only download and activate my software without touching Cursor(which a customer shouldn’t touch), allowing to have Cursor potential in a consumer app(maybe even mobile).