Skip to content
ARQCA

Windows desktop IDE

See what your code actually does. Then let AI help.

The AI-native IDE for people who inherited a codebase, not a demo.

Coming soonJoin early access

ARQCA is currently in development. There is no public installer yet — this page will not invent one.

Call path · illustration

Illustrated call path for POST /checkoutA route leads to a handler, then a service, a database commit, a payments call, and a heuristic email edge.POST /checkoutcheckoutHandlerOrderServicedb.commitpayments.chargenotify.email

The missing piece

AI can write code. But can it tell you what your code actually does?

Chat can write code. It cannot necessarily tell you whether that POST /checkout actually reaches the function you think it does.

When you are debugging a service you did not write, reviewing an AI patch you do not fully trust, or navigating an unfamiliar codebase, the missing piece is not more autocomplete.

It is ground truth.

ARQCA is a full IDE first — editor, terminal, git, search, tests, language intelligence — then AI on a map of the real program.

That's the product, not a plugin.

Features

Other tools help you type faster. ARQCA helps you see faster.

01

Plan before it writes

Ask for a change. See the architecture first: files, contracts, libraries, and the graph. Approve, refine, or discard. Nothing hits disk until you say so.

02

Follow the real call path

Pick an entry point — a route, a main, a test — and walk the cascade. Click a node to re-root. See callers. Know which edges are certain and which are heuristic.

03

Watch it run

Trace a request through the stack. Values show up on the line that produced them. Compare what you thought would happen with what actually did.

04

Chat that can see the graph

One chat. Attach files with @. Ground a question in a trace. Apply edits as diffs you can revert. Your model, your keys — cloud or local.

How it works

Plan → Code → Execute → Observe

The core development loop is the product's identity — not a side panel bolted onto an editor.

  1. 01

    Plan

    Decide the shape of the change. Understand architecture, files, and contracts first.

  2. 02

    Code

    Write it with the editor and agent under your rules. Review diffs before they land.

  3. 03

    Execute

    Run the real program on your machine. The website never becomes a proxy for that.

  4. 04

    Observe

    Watch the map and trace update so the next prompt is smarter than the last.

Who it's for

Built for developers working with real software.

  • Engineers inheriting a large or messy codebase
  • Backend teams debugging services and request flows
  • Leads who want architecture decisions to survive the next AI edit
  • Developers tired of “the model sounded confident”

Agentic development

The agent gets a map of the actual system. You keep the keys.

ARQCA gives the coding agent structured information about the software in front of you: the code graph, selected nodes, attached files, traces, and the prompt you typed. The model translates that map. It does not invent the architecture.

You choose the model in Settings. That can be Anthropic Claude, or another configured provider — including a local one. API keys stay on your PC. This website never collects them, never proxies them, and never runs your IDE in the browser.

  1. 01Program understanding
  2. 02Code graph + runtime trace
  3. 03AI context
  4. 04Your configured model
  5. 05Plan / agent
  6. 06Human approval
  7. 07Code diff

Stop guessing what the code does.

ARQCA is currently in development. Join the early-access list and we will send the Windows installer when it exists — not a placeholder.

Download for Windows — not yetJoin the early-access list