Editor
Works in VS Code, Cursor, and your favorite IDE.
Ask CodeNeter to plan or build anything
AgentClaude Opus 4.7
Build with AI.Without giving up the wheel.
CodeNeter is AI coding for your favorite editor and terminal — without locking you into one model or one way of working.
codeneter — agent
New Chat
migrate test runners from jest to vitest
codeneter — cli
explain how this codebase handles auth
/ commands · @ files · ! shell · ctrl+o output
Turn ideas into shipped code.
CodeNeter explores your repo, builds a clear plan, and handles implementation. You stay focused on what to build.
codeneter — plan mode
New Chat
add a global command palette with fuzzy file search
command-palette-plan.mdPending
Plans > command-palette-plan.md
Command palette
A modal overlay for fuzzy file search and quick actions, opened from anywhere with ⌘K.
Trigger
⌘K opens the palette anywhere in the app. Arrow keys navigate, Enter selects, Esc closes.
View Behavior
Modal with debounced fuzzy ranking across files, commands, and recent results.
3 Tasks
Add CommandPaletteView.tsx with results list
Register ⌘K in KeyboardManager.ts
Build fuzzy index over files and commands
Add a task, ⌘K to generate…
Wherever you code.
One agent across your editor, terminal, and any browser.
CLI
Run agents in any terminal, script, or session.
$ codeneter run▸ Reading workspace…▸ 3 tool calls queuedapprove · session · deny
Install Remote
Attach to a running CLI session from any browser — laptop or phone.
⌘ remote.codeneter.dev● live · session #f24a▸ streaming output…approve from phone
Open Built to do real engineering.
Everything the agent needs to ship real work, end to end — and everything you need to trust it.
Understands your codebase
Search across your repo, trace how code connects, and bring in exactly what matters.
Search "rate limit"
handlers.ts
auth.ts
3 files · ready to planBuilt to ship
Reads your code, makes changes, runs checks — and keeps you in the loop.
$ pnpm test rate-limit
allow deny
Use any model
GPT 5.5, Claude Opus 4.7, Kimi 2.6, GLM 5.1 — choose the right model for every task.
AutoBest for task
Claude Opus 4.7
GPT 5.5
+ Kimi 2.6, GLM 5.1, and more
Works in parallel
Multiple sessions on the same project, running in their own space.
● session-1 · feat/rate-limit● session-2 · refactor/checkoutisolated worktrees · no cross-talk
Efficient with every token.
CodeNeter handles large tool and MCP outputs without flooding your conversation, and loads MCPs only when you need them — so the agent stays fast and your context stays clean.
Smart output handling
Context24% / 200k
Only the useful slice of a big tool return reaches the model. The full output is kept on the side, ready when you need it.
Lazy MCP loading
On demandEach MCP server connects only when the agent reaches for it — no upfront cost, no flooded context.
githublinearpostgressentrystripeslackfsweb
MCPs load on demand, not all at once.
Start shipping with CodeNeter.
Pick any model · Credit rollover on paid plans