LangTailor Canvas

Visual LangGraph IDE — design agent workflows on a drag-and-drop canvas, then export clean, runnable Python.

LangTailor Canvas turns any *.langstitch.json file into a live visual editor for LangGraph applications. Drag in LLM nodes, tools, agents, RAG pipelines, guardrails, and routers; wire them together; and export a buildable, multi-module Python project — no lock-in, no rewrite.

It ships built into the LangTailor desktop IDE and installs into any VS Code–compatible editor (VS Code, VSCodium, Cursor, Gitpod, Theia) from Open VSX.

Features

• IR v2 documents — save graphs in the LangStitch representational layer (logical + presentation + target); v1 files migrate on load. Compile target is persisted in the document.
• SDK compiler build — Build and Run invoke langstitch compile for Python targets (with in-editor codegen fallback); dev runs enable RunEvents for local debugging.
• Visual graph editor — design LangGraph StateGraphs on an infinite canvas with drag-and-drop nodes, typed ports, and auto-layout.
• Rich node palette — LLM, Tool, Agent, RAG Agent, Multi-Intent Classifier, Decision, Function, and Subgraph nodes, plus your own custom components.
• Asset designers — author Skills, Guardrails, Business Rules, Personas, and RAG pipelines through dedicated forms; each exports to its own Python module.
• Custom Component Designer — build reusable nodes, connectors, and adaptors with a typed config schema and a safe codegen template (no eval, secrets emitted as os.environ.get(...)).
• Python 3.13 export — generate a clean multi-module project via the langstitch-sdk IR compiler (langstitch compile) or the built-in TypeScript codegen fallback — same scaffold the SDK CLI produces.
• Native editor integration — built as a VS Code custom editor, so undo/redo, dirty-state, and save all work exactly as you expect.
• Marketplace, in-editor — browse and add community plugins, personas, prompts, connectors, and .langstitch-pack.zip packs without leaving the canvas, or publish your own at marketplace.langstitch.com.
• Two project types — Agentic Development (graphs, skills, deploy) and Plugin Creator (multi-platform packs via langstitch plugin new / langstitch pack in the SDK CLI).
• Sign in with GitHub — authenticate to LangStitch right from the editor (secure browser-based OAuth) to sync your plugins to the LangTailor desktop IDE.
• Local toolchain — Build, Run, Test, Debug, Version, and Deploy run on your machine with keyboard shortcuts (Ctrl/Cmd+Alt+B/R/T/D/V and Ctrl/Cmd+Alt+Shift+D for deploy). No server required.

Getting started

1. Install the extension from Open VSX, or use the LangTailor desktop IDE where it's built in.

   # VSCodium / CLI
   codium --install-extension langstitch.langtailor-canvas
   

2. Create a graph — run LangTailor: New Graph from the Command Palette, or open any existing *.langstitch.json file.

3. Design your agent workflow on the canvas — add nodes, configure them in the property panel, and wire the edges.

4. Export to Python from the Platform panel and run it:

   pip install "langstitch-sdk[all]"
   python -m app
   

The LangStitch SDK

Exported projects run on the open-source LangStitch SDK — Python decorators, typed config, and a CLI for building LangGraph apps:

pip install langstitch-sdk
langstitch new my-agent

Docs: langstitch.com/docs · PyPI: langstitch-sdk

Links

• Website: langstitch.com
• Desktop IDE: langtailor.langstitch.com
• Marketplace: marketplace.langstitch.com
• Documentation: langstitch.com/docs
• Issues: github.com/LangStitch/langtailor/issues

Building from source

The canvas source is vendored in _canvas/ — no separate checkout needed.

npm install        # install build deps
npm run build      # build the webview bundle (_canvas/src) + extension host
# then press F5 in VS Code to launch the "LangTailor Canvas Extension" host

Open or create a *.langstitch.json file to launch the canvas editor.

To cut a release: bump version in package.json and desktop/package.json, tag langtailor-v<version>, and push (or run LangTailor Release manually). See docs/MAC_RELEASE.md for macOS signing and download-site updates. Open VSX: docs/OPENVSX_SETUP.md.

License

MIT © LangStitch