Is architecture-canvas free to use? What license does it have?

architecture-canvas is open source under the Apache-2.0 license. Check the license terms before using it in commercial products.

Is architecture-canvas still maintained?

Actively maintained. The last commit to architecture-canvas was on Oct 3, 2026.

What can I use architecture-canvas for?

Turning a requirements document into a first-draft system architecture diagram. Iterating on an existing diagram through conversation while keeping manual layout. Producing editable architecture diagrams from local files without sending them to a hosted editor. Adding document-reading and diagram-rendering tools to an AI agent over MCP.

How do I install architecture-canvas?

Clone the repository: git clone https://github.com/weike-zhang/architecture-canvas.git Enter the project directory: cd architecture-canvas Start the app (first run installs dependencies and builds the page): python3 start.py Open the workspace in your browser and configure a model in 模型设置

Where can I get architecture-canvas?

The source code and installation instructions are on GitHub: https://github.com/weike-zhang/architecture-canvas

weike-zhang/architecture-canvas

Local-first web workspace where you chat with uploaded documents to create and refine editable architecture diagrams on an Excalidraw canvas.

Open-source alternative to:Lucidchart draw.io

View on GitHub

Stars: 160
Forks: 12
Open issues: 0
Last commit: 17 days ago

What is it?

Architecture Canvas lets you upload requirement documents, spreadsheets, PDFs and images, then discuss them with an AI assistant that drafts an architecture diagram. The diagram lives on an editable Excalidraw canvas, so you can drag nodes, edit text and export SVG or PNG after the model produces a first draft.

Changes requested through chat are listed before they are applied, and fixed components can keep their positions. Each session keeps its own messages, documents, current diagram and generation history. The service runs on 127.0.0.1 only, and the model is reached through any Chat Completions-compatible endpoint with your own API key.

Manual editing and export work offline with the included demo session. Chat-based generation requires a configured model service that supports tool calls, and optional Tavily integration adds web search. The project is an Alpha release and includes a stateless Python SDK and an MCP stdio server.

Key features

01Generates architecture diagrams from uploaded documents, spreadsheets, PDFs and images through chat.
02Editable Excalidraw canvas with drag, text editing, undo and import.
03Exports diagrams as SVG, PNG and editable files that include images.
04Reviews proposed changes before applying them and can pin selected components.
05Runs locally on 127.0.0.1 with per-session data stored in a workspace folder.
06Provides a Python SDK and MCP stdio server for embedding the renderer and document reader.

Use cases

Turning a requirements document into a first-draft system architecture diagram.
Iterating on an existing diagram through conversation while keeping manual layout.
Producing editable architecture diagrams from local files without sending them to a hosted editor.
Adding document-reading and diagram-rendering tools to an AI agent over MCP.

Quick start

1
Clone the repository
git clone https://github.com/weike-zhang/architecture-canvas.git
2
Enter the project directory
cd architecture-canvas
3
Start the app (first run installs dependencies and builds the page)
python3 start.py
4
Open the workspace in your browser and configure a model in 模型设置

Things to know

Alpha version that runs locally only, bound to 127.0.0.1 with no multi-user accounts.
Chat, document understanding and diagram generation require a model service with an API key that supports tool calling.
Generation is limited to roughly 22 components and does not cover sequence, ER or complex branching diagrams.

Topics

#architecture-diagram #excalidraw #ai-agent #local-first #mcp #llm #diagram #tavily

FAQ

What is architecture-canvas?: Local-first web workspace where you chat with uploaded documents to create and refine editable architecture diagrams on an Excalidraw canvas. Architecture Canvas lets you upload requirement documents, spreadsheets, PDFs and images, then discuss them with an AI assistant that drafts an architecture diagram. The diagram lives on an editable Excalidraw canvas, so you can drag nodes, edit text and export SVG or PNG after the model produces a first draft.
Is architecture-canvas free to use? What license does it have?: architecture-canvas is open source under the Apache-2.0 license. Check the license terms before using it in commercial products.
Is architecture-canvas still maintained?: Actively maintained. The last commit to architecture-canvas was on Oct 3, 2026.
What can I use architecture-canvas for?: Turning a requirements document into a first-draft system architecture diagram. Iterating on an existing diagram through conversation while keeping manual layout. Producing editable architecture diagrams from local files without sending them to a hosted editor. Adding document-reading and diagram-rendering tools to an AI agent over MCP.
How do I install architecture-canvas?: Clone the repository: git clone https://github.com/weike-zhang/architecture-canvas.git Enter the project directory: cd architecture-canvas Start the app (first run installs dependencies and builds the page): python3 start.py Open the workspace in your browser and configure a model in 模型设置
Where can I get architecture-canvas?: The source code and installation instructions are on GitHub: https://github.com/weike-zhang/architecture-canvas

Summary generated by AI from the project's README. GitHub stats refresh daily.