session ~/zaguan-blade live
$ The AI editor with a map of your code

The AI that reads less but understands more.

contextSymbol-Indexed Context
idle~320 MB Idle
controlYou Review Every Diff

Zaguán Blade builds a live symbol index of your codebase, so the AI investigates before it answers — navigating by structure (functions, references, impact) instead of dumping files into context. That evidence-first approach pays off everywhere: sharper answers, lower bills on hosted models, and the difference between "toy" and "usable" for local ones. Every change lands as a diff you accept or reject.

Stuck on a visual bug? Capture a window, annotate the screenshot, and show the AI instead of describing it.

evidence zblade.session.mp4

Why Another Editor
Exists

Most AI editors are forks of the same heavyweight base, shipping your entire project as context and hoping the model guesses right. Blade was built from scratch to do the opposite — for work where a generic AI assistant isn't trustworthy enough.

The UI is separate from the AI daemon, so the editor stays fast. Context is evidence, not a dump: the AI gathers what's relevant, reasons from it, and shows its work — which means lower token costs and answers grounded in your actual code. Git is the source of truth, and every AI edit is a diff you approve. No telemetry. Run local models via Ollama and your code never leaves your machine.

Used daily to develop Blade and zcoderd. Tested on real production projects — not just demos.

The Lightweight Difference measured

Engineers trust metrics. Here is what idle resource usage looks like:

Tool Idle RAM
Zaguán Blade ~320 MB
VSCode + AI ~1.9 GB
Windsurf ~2.8 GB

Not for Everyone

  • If you want a VSCode clone, this isn't it.
  • If you need heavy UI plugins, this isn't it.
  • If you want lightweight, evidence-grounded AI workflows you control — this is for you.

What It's Actually
Good At

What It's Actually Good At

  • + Refactoring across multiple files.
  • + Showing the AI a screenshot instead of describing the bug.
  • + Reviewing AI patches safely with diff control.
  • + Getting sharp answers from local and hosted models with curated context.

Typical Flow in Blade

  1. Ask for a change.
  2. AI investigates: symbols, references, impact.
  3. A scoped diff appears.
  4. You review the change — and the evidence behind it.
  5. Accept or revert via Git.

No vibes-based edits. Evidence in, scoped change out — every step visible.

Zaguán Blade is under heavy development. Expect rapid updates, occasional breaking changes, and a fast feedback loop as we build toward v1.0.0.

1

Zaguán Blade (UI)

Active

Tauri v2 + Vite + React · User interface only. · Lightning fast.

2

zcoderd (AI daemon)

Active

Go server · State management · Tool execution & AI

3

Model Provider

Active

Zaguán AI: one subscription, every frontier model. Or bring your own Ollama / local endpoint.

Simple
Pricing

The Editor
Free to download and use
AI Features
Zaguán AI subscription — or bring your own local models via Ollama
Multi-model routing Priority support (email) Early access to new tools Funding development
View Pricing & Plans

Get Blade
for Your OS

! Pre-v1.0 software — breaking changes will happen. Hosted AI models require a Zaguán AI subscription; local models via Ollama work without one.