The crazy thing is nobody knows exactly who’s behind it.
But it’s absolutely huge.
A brand new coding model built for the way devs actually work today — agents, terminals, long sessions, and even images.
Open Cline or Windsurf or Cursor and you will see the new option hiding in your model picker:
code-supernova.
I previously thought it was a new native Windsurf model but no — this is a secret partner working with all the major IDEs.
Think of code-supernova as a tireless senior engineer who never forgets context. Its context window clocks in around 200,000 tokens—large enough to swallow your repo, your logs, your test output, your onboarding docs, and still have room for much more.
No more prompt-chopping. Your agent can hold the entire picture in its head and keep pushing forward.
And it’s multimodal. Feed it screenshots of a broken UI, an architecture diagram, a flowchart you sketched on paper, or a whiteboard photo from last night’s brainstorm.
It turns visual signals into code moves.
That unlocks a new workflow: design → diagram → implementation, without the translation overhead. Visual debugging becomes real. “Here’s the stack trace and the screenshot—fix it” becomes a single instruction, not a meeting.
Where code-supernova truly flexes is the agentic work.
This model is built for long-running sessions where your IDE agent plans, edits, runs tools, reads terminal output, evaluates results, and loops until the task is done. A true coding partner.
It reasons across files, updates the right modules in the right order, and keeps the terminal state in mind as it iterates. Refactors that used to take an afternoon shrink to a coffee break.
Cross-cutting changes stop feeling risky because the agent isn’t flying blind—it remembers every decision it just made.
The best part is it’s free to try right now in Cline, Kilo Code, Cursor, and Windsurf.
No new account. No platform migration. Just pick “code-supernova,” hand it a mission and watch your IDE light up.
It’s an alpha from a stealth lab, which makes this the perfect moment to build unfair advantage: learn its strengths, wire it into your flow, and ship faster while everyone else is still reading about it.
Privacy-conscious about sending your data to people you don’t know?
Flip through your IDE’s data-sharing settings and tune them how you like.
code-supernova is a real force multiplier: it’s built for how we really code—context-heavy, agent-driven, multimodal, and terminal-native.
Your model sees the whole board, think multiple moves ahead, and execute without losing the thread.
If you live inside an IDE agent and run long sessions, this is the model to beat.
Fire it up while it’s still free, point it at the messy, sprawling, real-world work on your plate, and let it run.