Copilot and Cursor both had a heavy July, and the things they shipped were not substitutes for each other.
That makes the comparison clearer than it has been for a while. The two tools are optimising for different problems.
What Copilot spent July building
Four new models from four vendors, agent skills and MCP in code review reaching GA on 29 July, default model enablement on the same day, team-level model policy on 31 July, and the August deprecation list.
Read as a whole, that is a platform maturing for organisations: control over who can use what, an automated reviewer that can consult internal systems, and a managed model lifecycle.
What Cursor spent July building
Side chats and conversation search on 10 July, Slack improvements on 17 July, Router on 22 July, an India-specific plan on 28 July, iPad on 29 July, and Google Workspace plugins on 3 August.
That is a tool pushing outward — to more devices, more places where work is discussed, and more price points.
How to choose
- Choose Copilot if governance is a real constraint — regulated data, multiple teams with different rules, or an organisation where an unreviewed model vendor appearing in the toolchain is a problem
- Choose Cursor if cost per request matters more than policy, or if your reviewers and decision-makers are away from a desk more than they are at one
- If neither of those is binding, the model list is now nearly identical and the decision comes down to which editor your team prefers
The honest caveat
This comparison has a shelf life measured in weeks. Both tools shipped six or more meaningful changes in July alone. Re-check before you commit a whole team.