d2-mcp
Create, validate, and render diagrams from D2 (Declarative Diagramming) code into SVG and PNG formats.
d2-mcp
A Model Context Protocol (MCP) server for working with D2: Declarative Diagramming, enabling seamless integration of diagram creation and validation into your development workflow.
Tools:
- Compile D2 Code
- Validate D2 syntax and catch errors before rendering
- Get immediate feedback on diagram structure and syntax
- Accepts either direct code or file path to D2 file
- Render Diagrams
- Generate diagrams for visual feedback and refinement
- Support PNG, SVG, and ASCII output formats
- Accepts either direct code or file path to D2 file
- Fetch D2 Cheat Sheet
- Returns a Markdown reference covering shapes, styling, and transport usage
Install
Option 1: Install Binary Release
Option 2: Install via go
go install github.com/h0rv/d2-mcp@latest
Option 3: Build Locally
git clone https://github.com/h0rv/d2-mcp.git
cd d2-mcp
go build .
Option 4: Build Image Locally
docker build . -t d2-mcp
# Run in stdio mode (default - for MCP clients)
docker run --rm -i d2-mcp
# Run in stdio mode with filesystem access
docker run --rm -i -v $(pwd):/data d2-mcp
# Run in SSE mode (HTTP server)
docker run --rm -e SSE_MODE=true -p 8080:8080 d2-mcp
# Run in SSE mode with filesystem access
docker run --rm -e SSE_MODE=true -p 8080:8080 -v $(pwd):/data d2-mcp
Option 5: Run Container Image
# Run in stdio mode (default - for MCP clients)
docker run --rm -i ghcr.io/h0rv/d2-mcp:main
# Run in stdio mode with filesystem access
docker run --rm -i -v $(pwd):/data ghcr.io/h0rv/d2-mcp:main
# Run in SSE mode (HTTP server)
docker run --rm -e SSE_MODE=true -p 8080:8080 ghcr.io/h0rv/d2-mcp:main
# Run in SSE mode with filesystem access
docker run --rm -e SSE_MODE=true -p 8080:8080 -v $(pwd):/data ghcr.io/h0rv/d2-mcp:main
Setup with MCP Client
MacOS:
# Claude Desktop
$EDITOR ~/Library/Application\ Support/Claude/claude_desktop_config.json
# OTerm:
$EDITOR ~/Library/Application\ Support/oterm/config.json
Add the d2 MCP server to your respective MCP Clients config:
Using Binary:
{
"mcpServers": {
"d2": {
"command": "/YOUR/ABSOLUTE/PATH/d2-mcp",
"args": ["--image-type", "png"]
}
}
}
Using Binary with file output:
{
"mcpServers": {
"d2": {
"command": "/YOUR/ABSOLUTE/PATH/d2-mcp",
"args": ["--image-type", "png", "--write-files"]
}
}
}
Using Docker:
{
"mcpServers": {
"d2": {
"command": "docker",
"args": ["run", "--rm", "-i", "ghcr.io/h0rv/d2-mcp:main", "--image-type", "svg"]
}
}
}
Using Docker with filesystem access:
{
"mcpServers": {
"d2": {
"command": "docker",
"args": [
"run", "--rm", "-i",
"-v", "./:/data",
"ghcr.io/h0rv/d2-mcp:main",
"--image-type", "ascii",
"--ascii-mode", "standard",
"--write-files"
]
}
}
}
Rendering Formats
The server returns PNG output by default when rsvg-convert from librsvg is available. If rsvg-convert is missing, the server automatically removes png from the render-d2 tool's format enum and falls back to SVG. The Docker image installs librsvg plus fontconfig/DejaVu fallback fonts, so PNG works out of the box there.
Install librsvg for local binaries:
# macOS
brew install librsvg
# Debian/Ubuntu
sudo apt-get install librsvg2-bin
# Alpine
apk add librsvg
Override globally when starting the binary:
./d2-mcp --image-type svg # SVG output
./d2-mcp --image-type ascii # ASCII output with Unicode box drawing characters
./d2-mcp --image-type ascii --ascii-mode standard # ASCII output restricted to basic ASCII chars
Inside MCP tool calls, pass the optional format argument (png, svg, ascii) and, when ascii, the ascii_mode argument (extended, standard) to switch formats per request.
Docker Usage Examples
Run the container with default PNG output over stdio:
docker run --rm -i ghcr.io/h0rv/d2-mcp:main
Switch to Unicode ASCII diagrams and capture responses as plain text:
docker run --rm -i ghcr.io/h0rv/d2-mcp:main --image-type ascii
Use basic ASCII characters and write rendered files back into your working tree (requires a bind mount):
docker run --rm -i \
-v "$(pwd)":/data \
ghcr.io/h0rv/d2-mcp:main \
--image-type ascii \
--ascii-mode standard \
--write-files
Expose the SSE server on port 8080 while emitting SVG:
docker run --rm -e SSE_MODE=true -p 8080:8080 ghcr.io/h0rv/d2-mcp:main --image-type svg
Expose the streamable HTTP transport (default endpoint /mcp) for use with MCP clients that expect the new protocol:
docker run --rm -p 8080:8080 ghcr.io/h0rv/d2-mcp:main --transport http --image-type svg
Cheat Sheet Tool
Retrieve the built-in quick reference as Markdown:
{
"tool": "fetch_d2_cheat_sheet"
}
The cheat sheet highlights common shapes, layout tips, and ASCII-friendly patterns, making it ideal support material for downstream LLM prompts.
Transports
The server defaults to stdio transport for CLI-driven MCP clients. Switch transports per run:
--transport stdio: default for local CLI integrations.--transport sse: legacy Server-Sent Events transport (alias:--sse).--transport http: streamable HTTP transport on/mcp; combine with-p/--portwhen running in Docker or containers.
Environment overrides:
MCP_TRANSPORTsets the transport (stdio,sse,http) when flags are not provided.PORT(or the legacySSE_PORT) sets the listening port for SSE/HTTP transports.SSE_MODE=trueretains backwards compatibility by selecting the SSE transport.
Tool Reference
| Tool | Description | Key Arguments |
|---|---|---|
compile-d2 | Validates D2 source and surfaces syntax errors. | code (string) or file_path (string) |
render-d2 | Renders diagrams to PNG, SVG, or ASCII (ASCII is LLM-friendly). | code/file_path, format (png, svg, ascii), ascii_mode (extended, standard) |
fetch_d2_cheat_sheet | Returns a Markdown cheat sheet with examples and best practices. | None |
Tip: Run compile-d2 first to validate, then call render-d2 with the same payload for the final output.
Development
Debugging
npx @modelcontextprotocol/inspector /YOUR/ABSOLUTE/PATH/d2-mcp/d2-mcp
관련 서버
Alpha Vantage MCP Server
스폰서Access financial market data: realtime & historical stock, ETF, options, forex, crypto, commodities, fundamentals, technical indicators, & more
Loop MCP Server
Enables LLMs to process array items sequentially with a specific task.
VoteShip
MCP server for VoteShip - manage feature requests, votes, roadmaps, and changelogs from any MCP client. 22 tools, 5 resources, 4 workflow prompts. Triage feedback, detect duplicates, plan sprints, and generate changelogs with AI.
Remote MCP Server (Authless)
An example of a remote MCP server deployable on Cloudflare Workers without authentication.
Chronos Protocol
A robust MCP server that eliminates temporal blindness in AI coding agents through intelligent time tracking, persistent memory, and complete session traceability.
Kubernetes Automated Installation
An agent for automatically installing Kubernetes in a Rocky Linux environment using MCP.
DeepWiki by Devin
Remote, no-auth MCP server providing AI-powered codebase context and answers
widemem.ai
Open-source AI memory layer with importance scoring, temporal decay, hierarchical memory, and YMYL prioritization
depwire
Code dependency graph and AI context engine. 10 MCP tools that give Claude, Cursor, and any MCP client full codebase context — impact analysis, dependency tracing, architecture summaries, and interactive arc diagram visualization. Supports TypeScript, JavaScript, Python, and Go.
codegraph-rust
A blazingly fast codebase graphRAG implementation in 100% Rust
MCP ZAP Server
Exposes OWASP ZAP as an MCP server, enabling AI agents to orchestrate security scans, import OpenAPI specs, and generate reports.