Add Autonomous “Goal Mode” Similar to Claude Code’s /goal

Ideas Feature Requests
1

Feature request for product/service

Cursor IDE

Describe the request

Claude Code recently introduced a /goal feature that enables true autonomous multi-turn execution inside the coding workflow.

Example:

/goal all authentication tests pass and lint is clean

The agent continues working across multiple turns automatically until the completion condition is satisfied.

This is significantly more powerful than simple chat persistence or custom instructions because it introduces:

completion-condition driven workflows
autonomous continuation between turns
iterative self-correction
evaluator-based stopping conditions
unattended long-running development tasks

This would be an incredible addition to Cursor.

Suggested behavior:

Define a goal for the current session/workspace
Cursor agent keeps iterating until:
tests pass
build succeeds
acceptance criteria are met
issue checklist is completed
Automatic progress evaluation after each iteration
Clear visibility of:
current goal
iterations performed
why execution continues/stopped
Optional limits:
max iterations
token budget
timeout
Support for autonomous refactors and migration tasks

Example use cases:

“Migrate project to React 19 until build passes”
“Fix all TypeScript errors without modifying generated files”
“Refactor auth module until all tests succeed”
“Resolve all ESLint issues in /frontend”

This would push Cursor much closer to a true agentic development environment instead of a turn-by-turn assistant.

3

More Example Use Cases

Frontend

  • “Upgrade Next.js app to App Router until production build passes”

  • “Fix all hydration issues in the React app”

  • “Migrate Tailwind v3 → v4 without breaking UI”

  • “Remove all deprecated Vue APIs and ensure tests pass”

Backend

  • “Refactor Express API to Fastify while keeping all integration tests green”

  • “Migrate REST endpoints to tRPC incrementally”

  • “Resolve all NestJS dependency injection errors”

  • “Optimize slow PostgreSQL queries until benchmarks improve”

TypeScript / Monorepo

  • “Eliminate all any types in /packages/core”

  • “Fix all TypeScript errors across the monorepo”

  • “Convert legacy JS modules to strict TypeScript”

  • “Standardize ESLint and Prettier configs across workspaces”

DevOps / Infrastructure

  • “Fix failing GitHub Actions workflows”

  • “Migrate Docker Compose setup to Kubernetes manifests”

  • “Reduce CI runtime below 10 minutes”

  • “Update Terraform modules until terraform plan is clean”

Mobile

  • “Upgrade React Native app to latest stable version”

  • “Fix all Android build failures”

  • “Migrate Expo SDK to latest version”

  • “Resolve iOS Swift compatibility warnings”

Python / AI

  • “Refactor FastAPI project to async architecture”

  • “Fix all mypy typing issues”

  • “Optimize PyTorch training pipeline memory usage”

  • “Update LangChain implementation to latest APIs”

Java / JVM

  • “Migrate Spring Boot 2 → 3 until all tests pass”

  • “Upgrade Gradle and fix build incompatibilities”

  • “Replace deprecated Hibernate APIs”

  • “Modularize legacy Java services”

Go

  • “Refactor project to context-aware request handling”

  • “Eliminate data races detected by go test -race”

  • “Migrate REST handlers to gRPC”

  • “Improve benchmark performance without changing API behavior”

Rust

  • “Resolve all Clippy warnings”

  • “Refactor async runtime to Tokio”

  • “Remove unnecessary clones and allocations”

  • “Make project compile on stable Rust”

Security / Quality

  • “Fix all high severity npm audit issues safely”

  • “Add input validation to all public API endpoints”

  • “Resolve OWASP security findings”

  • “Increase test coverage above 80%”

Autonomous Multi-Step Tasks

  • “Analyze the repository, create a migration plan, execute it step-by-step, and stop only when CI passes”

  • “Continuously fix failing tests until the suite is green”

  • “Refactor duplicated business logic across services while preserving behavior”

  • “Upgrade dependencies incrementally and rollback problematic updates automatically”

5

Hey, thanks for the detailed feature request and the use case examples. It really helps us see what kind of UX you expect from autonomous mode.

Got it, the idea is a declarative goal plus autonomous iterations until a condition is met. While there isn’t an official Goal Mode yet, you can already put together part of this workflow using public Cursor features:

  • Skills Agent Skills | Cursor Docs You can define a custom skill that the agent triggers, like iterate until tests pass and lint is clean. This is the closest to a /goal-style UX.
  • Plan Mode Plan Mode | Cursor Docs For breaking the goal into steps before an autonomous run.
  • Cloud Agents Cloud Agents | Cursor Docs Autonomous execution in a VM, great for long migrations or refactors like in your examples.
  • Automations Automations | Cursor Docs Cron or webhook triggers if you need to rerun until you get the result.

It’s not a 1 to 1 match with what you described, but a lot of scenarios like framework migration, fix all TS errors, or driving tests to green are already possible by combining Plan Mode plus Cloud Agent plus a custom skill with a completion condition.

If other users also want this and the thread gets upvotes, it helps with prioritization. I can’t share an ETA.