IdexalGet early access

Features

Built to ship, not to autocomplete.

Idexal is a full agentic IDE: it plans, edits, runs, and verifies — across your whole workspace, on your machine.

The loop

Plan → write → run → verify

Idexal doesn't just autocomplete. It decomposes your goal into a plan, edits files, executes commands, reads the output, and self-corrects until the task actually works — showing every step as it goes.

Multi-step task planning
Real command execution
Reads results & self-corrects
Full, inspectable trace

Extend

100+ extensions via MCP

Equip the agent with capabilities through the Model Context Protocol — shell, browser, memory, databases, your internal APIs, and anything you can wrap in a tool.

Model Context Protocol
Built-in dev tools
Bring your own servers
One-click enable

Reuse

Recipes & Skills

Capture a workflow once and run it forever. Recipes are parameterized, shareable agent playbooks; Skills are domain capabilities you can equip in a click.

Parameterized recipes
Shareable & versioned
Domain skill packs
Schedule them to run

Models

Any model, your keys

Use Claude, GPT, Gemini, or local models. Switch providers without rewriting a thing. You bring the keys; nothing is marked up.

Anthropic · OpenAI · Google
Local & open models
No markup, your keys
Swap mid-project

Private

Local-first & cert-pinned

The engine runs on your machine over a certificate-pinned local connection. Your code, your keys, and your data stay with you — not on someone else's server.

Runs on your hardware
Cert-pinned local engine
No code leaves your box
No lock-in

Experience

An IDE built for agents

A focused, glass-dark command surface with a live activity feed, a command palette, sessions, and everything the agent is doing — visible and under your control.

Live activity feed
Command palette (⌘K)
Sessions & history
Cost & token visibility

Ready to build with an agent?

Free while in preview. Bring your own keys.

Get early access