An open source AI engineering environment for local AI and hosted Zaguán workflows.
Zaguán Blade is a desktop code editor built for AI-assisted software work: project exploration, codebase investigation, file editing, terminal workflows, debugging, and validation. Use it with local models for private work, or connect to the hosted Zaguán Coder Daemon for the full power experience.
Blade itself is open source and MIT licensed. You can build it, inspect it, modify it, and use it with local AI without a subscription.
The advanced Zaguán workflows are powered by the proprietary Zaguán Coder Daemon, a subscription-backed backend for deeper AI orchestration.
Get a subscription →Local AI is a first-class part of Blade. You can explore projects, ask questions, and perform lighter AI-assisted tasks without a subscription. Local model quality varies, and some workflows require stronger hosted models or backend orchestration.
Read the local AI setup docs at zblade.dev/docs →Inspect, gather evidence, run approved tools, make changes, track what changed, and preserve context. Built for real engineering work where a confident answer is not enough.
Tree-sitter parsing, symbol indexing, file operations, and context assembly all run in native Rust. 10× faster operations and 5× lower memory than JavaScript-based editors.
AI writes directly to disk with history snapshots. File watcher reloads automatically. Accept keeps the change; reject restores from snapshot. No preview-vs-actual confusion.
Blade Protocol handles WebSocket communication with the hosted daemon; BCP manages internal IPC with versioning, domains, intent/event causality, and idempotency.
Symbol-based selection, token budget management, semantic relationships, and parallel processing give the AI exactly what it needs—lower costs, better results.
Loop detection, stagnation detection, tool-spam prevention, and parallel read optimization prevent runaway tokens and wasted execution.
Automatic snapshots before any modification, an uncommitted changes panel, per-file accept/reject with diff previews, and full group undo history.
Built with Tauri v2, not Electron. Compact 12MB packages, 50–100MB memory usage, <100ms startup, and native Rust backend speed without Chromium overhead.
React 19 + Vite, CodeMirror 6, xterm.js, TailwindCSS 4. Minimal UI logic; dispatches intents via BCP and listens for events.
Tauri v2 + Tokio. Blade Protocol dispatcher, AI workflow engine, tree-sitter parser, symbol index, history, file watcher, and Git operations.
Hosted proprietary AI backend via WebSocket. Context optimization, multi-model support, tool orchestration, and subscription-backed power features.
Download pre-built binaries for Windows, macOS, and Linux from GitHub releases. Or build from source with Rust and Bun.
Requires Rust 1.70+, Bun 1.3+, and platform dependencies.
git clone https://github.com/ZaguanLabs/ZaguanBlade.git
cd ZaguanBlade
bun install
bun run tauri build
# Dev server
bun run tauri devLocal model capability varies. Most power features require a Zaguán subscription. See zblade.dev/docs for setup and usage guidance.
Zaguán Blade is not trying to be a VS Code fork with a chat panel. It is built around a few practical beliefs:
Developers should be able to inspect, modify, and build the desktop environment they use.
Users should be able to try Blade and run useful workflows without subscribing first.
AI changes should be visible to the filesystem, Git, tests, and other tools immediately.
The AI should inspect, search, run tools, validate, and leave a trail instead of simply producing confident text.
Active Development: Zaguán Blade v0.8.2 is near-stable and very close to production-ready. Expect frequent updates and a fast feedback loop while the editor and daemon mature.
MIT Licensed: Zaguán Blade is free and open source. The hosted Zaguán Coder Daemon requires a subscription, but the editor itself is yours to use, modify, and distribute.
Changelog: zblade.dev/changelog
Bug reports, feature suggestions, PRs, docs, and translations are welcome on GitHub.
Visit the repository →Online docs, user guide, and build instructions at zblade.dev and in the repo docs/ directory.
Read the docs →