I built CortexIDE as a simple alternative to Cursor, mainly because the open-source editor Void (which CortexIDE is based on) slowed down, and I still wanted a local-first, hackable AI editor I could control.
CortexIDE is a VS Code–based editor, not an extension. The workflow is intentionally explicit: chat with the model, generate a concrete plan, produce diffs, and apply changes only after review.
It supports chat, inline edit, and agent-style multi-file edits, repo-aware context via indexing, local models (e.g. Ollama) as well as remote providers, PDF uploads for specs/docs, image uploads for visual context, checkpoints/rollback, todo list for large task and no telemetry by default.
This is still early and actively evolving. Performance on large repos and slower machines is something I’m actively improving.
I’m sharing this to get feedback from people who like Cursor’s UX but want more control, prefer local or self-hosted models, or don’t trust extensions to freely mutate their repo.
Pterjudin•1mo ago
I work on large codebases during the day, and I built this at night because I wanted a local-first AI editor I could actually control. Extensions worked fine for small edits, but once I needed multi-file changes, reviewable diffs, and predictable behavior, things started to break down.
CortexIDE is a VS Code–based editor focused on explicit AI edits: plan first, show diffs, then apply only after review. It supports local models, repo-aware context, PDF/image uploads for additional context, and checkpoints so changes can be rolled back.
This is still early and I’m actively improving performance and reliability, especially on larger repos.
I’d really appreciate feedback on where AI editors fail for you, what you don’t trust them with, and what would make you switch (or not).