Feature Request: Goal-Pursuing Autonomous Agent Mode

Ideas Feature Requests
,
1

Feature request for product/service

– Other –

Describe the request

I’d love Cursor to support a goal-pursuing autonomous mode, where it can work independently toward completing a task until it is actually finished.

Instead of:

Prompt → Response → Wait

Cursor would work in a loop:

Goal → Plan → Execute → Verify → Iterate → Complete

The Idea

Users give Cursor a high-level objective:

“Implement Google OAuth.”

or

“Fix all TypeScript errors and make CI pass.”

Cursor then:

Creates a plan
Breaks work into subtasks
Executes changes
Runs tests
Debugs failures
Adapts strategy
Continues working until completion or blocked

This feature would introduce a persistent autonomous workflow, where Cursor actively pursues an end goal over many steps.

Verification:

Before saying a task is complete, Cursor should verify results by:

Running tests
Checking builds
Validating logs
Confirming requirements were met

4

hey @rat_rad, have you tested cloud agent long running mode? what’s missing from them for you?

6

Hey, thanks for the feature request. There’s already a similar thread here: Add Autonomous “Goal Mode” Similar to Claude Code’s /goal: Add Autonomous “Goal Mode” Similar to Claude Code’s /goal. It makes sense to add your vote and comments there so the signal doesn’t get diluted.

On the request itself, Cloud Agents and Plan Mode already partly cover this use case like planning, autonomous execution, and running tests, but without a formal goal completion contract with verification and iterating until the result. Like Tom asked above, can you share what exactly is missing in the current Cloud Agent long-running mode? That’ll help us understand if we really need a separate mode, or if we should just tune the existing one.

Helpful links: