The AI-native code editor.

Q: How does Vortex handle pricing?

Vortex is free to get started. You bring your own API key for cloud models (OpenAI, Anthropic, OpenRouter), and those costs go directly to the provider. Local models run at no cost on your own hardware — no API key required.

Q: What AI models does Vortex support?

Vortex supports OpenAI (GPT-4o, o3, etc.), Anthropic (Claude Sonnet, Opus), and OpenRouter — which gives access to hundreds of models from one API key. You can also point it at any OpenAI-compatible endpoint, including self-hosted models. For local inference, Vortex uses llama.cpp and can download GGUF models directly from Hugging Face.

Q: Can I use Vortex without an API key?

Yes. With local models you don't need any API key. Download a GGUF model in Vortex settings and the inference runs entirely on your machine — Metal-accelerated on Apple Silicon. No account, no API key, no data leaving your device.

Q: Is my code sent to any server?

Only if you use a cloud model. When you send a message, the relevant context goes directly to the provider you configured — OpenAI, Anthropic, etc. Vortex has no server in between. Codebase indexing is entirely local; your code never leaves your machine as part of the search index.

Q: How do I migrate from Cursor?

Vortex reads .cursorrules and .cursor/mcp.json out of the box. Any rules and MCP server configurations you set up in Cursor carry over automatically. Open your existing project folder in Vortex and it will pick up your configuration.

Q: Does Vortex work with my existing codebase?

Yes. Open any workspace folder and Vortex indexes it locally. The agent reads and edits your actual files — it doesn't need a clean repo or a specific framework. It works with Rust, TypeScript, Python, Go, or any language your terminal and tools can handle.

Q: What makes it different from a coding copilot?

Copilots suggest code in your current file. Vortex runs a full agent loop: it inspects your repo, creates a plan, edits multiple files, runs your tests, and hands back a reviewable summary — all in one session. It's the difference between an autocomplete and an engineer working alongside you.

Q: How fast can I get started?

Sign up for the closed beta to get early access. Once you're in, open a project folder and configure a model (local or cloud). You can start delegating tasks in under ten minutes. No team onboarding flow, no lock-in.

Plan, edit, verify, and ship — with local or cloud AI models, full repo context, and an agent that works autonomously while you stay in control.

Vortex code editor showing the file explorer, editor, and AI agent panel

Why Vortex Bring your own AI model · Run models locally · Approve every edit · .cursorrules & MCP · macOS, Windows, Linux

Capabilities

A full agent loop, built into a standalone editor.

Vortex is not a plugin or an autocomplete add-on. It's a complete AI-native editor with planning, multi-file editing, review mode, and verification built in.

Ask, Plan, and Agent modes: Switch between asking questions, creating structured plan files, and running a full autonomous agent loop — all from the same chat interface.
On-device codebase search: Your code is indexed locally with on-device semantic search and symbol tracing. Nothing is uploaded to a third-party indexing service.
Plan mode and review mode: Agent changes appear as pending diffs you can approve or reject one by one. Plan files in .vortex/plans/ let you edit intent before execution starts.
Multi-file editing with diffs: The agent patches files across your whole repository with structured tool calls. A built-in diff viewer shows exactly what changed and why.
Integrated terminal and verification: A full PTY terminal runs inside the editor. The agent can execute commands, run tests, read diagnostics, and act on the results.
MCP and Cursor compatibility: Vortex reads .cursorrules and .cursor/mcp.json. Connect any MCP server to extend the agent with custom tools and data sources.

Model flexibility

Bring your own model. Or run one locally.

Vortex is not tied to a single AI provider. Connect OpenAI, Anthropic, or OpenRouter — or download a GGUF model and run it entirely on your own hardware.

Local models via llama.cpp — Metal-accelerated on Apple Silicon. Download GGUF models from Hugging Face directly in Vortex settings.
OpenAI, Anthropic, OpenRouter — use any frontier model. Bring your own API key; traffic goes directly to the provider.
Any OpenAI-compatible endpoint — self-hosted models, local Ollama instances, or enterprise gateways.

llama.cpp

Runs entirely on your machine. No API key, no data leaving your device. Metal-accelerated on macOS.

Free · private

OpenAI

GPT-4o, o3, and the full OpenAI model lineup. Bring your own key.

Your key

Anthropic

Claude Sonnet, Opus, and future Claude models. Direct API access.

Your key

OpenRouter

Access hundreds of models from one API key. Or use any OpenAI-compatible endpoint.

Your key

How it works

From open project to reviewed, verified changes — in one loop.

Vortex works the way experienced engineers work: understand the codebase first, plan the approach, implement carefully, verify the result, and hand back something you can actually ship.

Step 1

Open your project and add context

Point Vortex at any workspace. It indexes your codebase locally. Use @ mentions to attach specific files, symbols, URLs, or git history to your message.

Step 2

Pick a mode and describe what you need

Choose Ask for a quick answer, Plan to generate a reviewable plan file before any code changes, or Agent to run the full implementation loop autonomously.

Step 3

Review diffs, run checks, and ship

Each file change appears as a pending diff. Approve or reject edits individually. The agent runs your tests and reports diagnostics before you ever touch git.

55+

Agent tools

Platforms

100% local

Code indexing & search

Closed Beta

Get early access to Vortex.

We're rolling out access to a small group of developers. Sign up and we'll let you know when it's your turn.

macOS, Windows, and Linux Local + cloud AI models Free to get started

Why Vortex

What you get that IDE copilots don't provide.

Most AI coding tools are extensions that augment an existing editor. Vortex is a standalone editor built ground-up for autonomous workflows — with more transparency, more control, and no vendor lock-in.

Run AI models locally

llama.cpp with Metal acceleration. No API key, no data leaving your machine.

IDE copilots: Cloud-only, no local option.

Choice of AI provider

OpenAI, Anthropic, OpenRouter, or any compatible endpoint. Direct API access.

IDE copilots: Typically routed through a vendor proxy.

Approve or reject each change

Review mode with pending diffs. Accept or discard edits one by one.

IDE copilots: Often applied automatically.

Structured plan files

Editable plan files in .vortex/plans/ before the agent runs.

IDE copilots: Plans shown in chat, not editable artifacts.

On-device codebase indexing

Local RAG — your code never leaves your machine for indexing.

IDE copilots: Code uploaded to cloud indexing services.

Standalone desktop app

Native Tauri app. No VS Code dependency, no Electron fork.

IDE copilots: Usually a VS Code fork or extension.

FAQ

Common questions from developers evaluating Vortex.

Everything you need to know before adding Vortex to your workflow.

How does Vortex handle pricing?

What AI models does Vortex support?

Can I use Vortex without an API key?

Is my code sent to any server?

How do I migrate from Cursor?

Does Vortex work with my existing codebase?

What makes it different from a coding copilot?

How fast can I get started?

Closed Beta

Your code, your models, your control.

Vortex is coming soon. Sign up for the closed beta and be among the first to build with an AI-native editor — no lock-in, no vendor dependency.

Join the Closed Beta