diag2md-mcp is an MCP server that bridges Draw.io C4/UML diagrams with AI coding agents by converting them to Mermaid Markdown.
list_diagrams: Discovers Draw.io architecture diagram files in the workspace using configurable glob patterns and ignore rules.
convert_diagrams_read: Batch-scans and converts diagrams into Mermaid Markdown, returning the content directly as AI context.
convert_diagrams_write: Batch-scans, converts, and writes updated .md files alongside the diagram sources for documentation synchronization.
Supports c4 and uml diagram types, with defaults configurable via DIAG2MD_TYPE and per-request diagramType overrides.
Serves as an architectural guardrail for AI agents, helping enforce C4 system and container boundaries before generating code.
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type @ followed by the MCP server name and your instructions, e.g., "@diag2md-mcp convert all Draw.io diagrams in the architecture folder to Mermaid markdown"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
Model Context Protocol (MCP) server providing AI coding agents with Draw.io C4 and UML architectural guardrails and diagram conversion capabilities.
diag2md-mcp bridges Draw.io architecture diagrams (.xml, .drawio) with AI coding agents (such as Antigravity, Cursor, Claude Desktop, and VS Code MCP clients) by converting C4 and UML diagrams into structured Mermaid Markdown in real time.
Powered by diag2md.
The Problem: AI coding assistants cannot natively interpret visual architecture diagrams (such as Draw.io files). Without a way to parse these diagrams, AI agents frequently generate code that violates a project's established C4 system and container boundaries.
The Impact: Attempting to solve this by pasting massive, static architectural documents into the AI's prompt severely bloats the context window and reduces the model's performance. As a result, projects suffer from "architectural drift," where the AI-generated codebase slowly misaligns with the intended system design and dependency rules.
The Need: There is a need for a dynamic, on-demand integration (via an MCP server) that translates visual Draw.io C4 diagrams into an AI-readable format (Mermaid Markdown). This solution will serve as a strict, real-time architectural guardrail, ensuring every AI prompt and code change remains aligned with the project's single source of truth.
Why the C4 Model?: The C4 model is the ideal architectural language for this bridge because it caters to both human and machine audiences perfectly. Its hierarchical structure (Context, Containers, Components, Code) provides high-level visual clarity that is easy for non-technical stakeholders and business leaders to understand. Simultaneously, its strict categorization provides the exact deterministic boundaries and structural logic that an AI needs to reason about system architecture, making it the perfect standard for AI-assisted engineering guardrails.
Related MCP server: drawio-mcp
AGENTS.md / .cursorrules)You can enforce strict architectural compliance across your team by adding an architectural rule directive to your project's AGENTS.md, GEMINI.md, or .cursorrules:
### ARCHITECTURAL RULES
Before writing any new modules, creating new services, or adding dependencies,
you MUST call the `convert_diagrams_read` MCP tool to verify that proposed changes
align with the C4 Draw.io architecture. Do not suggest structural changes that
violate these boundaries.Automated Workspace Discovery: Scans project directories for Draw.io diagram files matching configurable glob patterns (**/architecture/**/*.xml, **/*.drawio).
High-Performance In-Memory Conversion: Converts Draw.io XML models to Mermaid C4 & UML Markdown in-memory without subshell process spawning overhead.
AI Context Provider (convert_diagrams_read): Reads and converts diagrams dynamically to provide rich architectural context directly to AI coding assistants during chat sessions.
Batch Diagram Synchronization (convert_diagrams_write): Keeps architecture documentation up to date by generating .md files alongside diagram sources.
npxYou can run the MCP server directly without pre-installing:
npx -y diag2md-mcpnpm install -g diag2md-mcpTo connect diag2md-mcp to your favorite AI assistant or MCP client, add the server to your client configuration file (e.g., mcp_config.json, claude_desktop_config.json, Cursor, or Antigravity MCP settings).
{
"mcpServers": {
"diag2md-mcp": {
"command": "diag2md-mcp"
}
}
}{
"mcpServers": {
"diag2md-mcp": {
"command": "node",
"args": ["/path/to/diag2md-mcp/dist/index.js"]
}
}
}npx{
"mcpServers": {
"diag2md-mcp": {
"command": "npx",
"args": ["-y", "diag2md-mcp"]
}
}
}diag2md-mcp exposes 3 core tools to AI coding agents:
Tool Name | Description | Arguments |
| Discovers all Draw.io architecture diagram files ( |
|
| Batch scans workspace diagram files matching glob patterns, converts them, and writes updated |
|
| Batch scans workspace diagram files and returns converted Mermaid Markdown directly as context for the AI assistant. |
|
The MCP server settings can be customized via environment variables:
Environment Variable | Description | Default Value |
| Comma-separated glob patterns to discover diagram files. |
|
| Comma-separated glob patterns to ignore during file discovery. |
|
| Default conversion diagram type ( |
|
env PatternsYou can configure custom search patterns directly in your MCP server JSON configuration using the env block:
{
"mcpServers": {
"diag2md-mcp": {
"command": "diag2md-mcp",
"env": {
"DIAG2MD_PATTERNS": "**/architecture/**/*.xml, **/docs/**/*.drawio",
"DIAG2MD_IGNORE": "**/tmp/**, **/node_modules/**",
"DIAG2MD_TYPE": "c4"
}
}
}
}patternsAI assistants or tools can also override search patterns dynamically per request:
{
"name": "convert_diagrams_read",
"arguments": {
"patterns": [
"**/architecture/**/*.xml",
"**/docs/**/*.drawio"
]
}
}# Clone the repository
git clone https://github.com/diag2md/diag2md-mcp.git
cd diag2md-mcp
# Install dependencies
npm install
# Build TypeScript to JavaScript dist/
npm run build
# Run unit tests
npm run test
# Run type checker
npm run typecheckBatch scan and convert all workspace diagram files (.xml, .drawio) into Mermaid Markdown, returning their content directly as AI context.
| Name | Required | Description | Default |
|---|---|---|---|
| patterns | No | Custom glob patterns to scan for diagram files | |
| diagramType | No | Diagram type: "c4" or "uml" (default: "c4") |
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
With no annotations, the description carries the full behavioral burden. It usefully discloses batch behavior, file-type scope, output format, and that results are returned to the caller. However, it never explicitly states the operation is non-destructive/read-only (only the name's 'read' hints at it), and it does not warn that batch-converting 'all' workspace files could produce very large context output.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
A single sentence with zero filler: action and scope are front-loaded, followed by output format and return semantics. Every element earns its place, and no information is repeated from the schema.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
For a tool with only 2 optional parameters, no output schema, and no annotations, the description supplies the core facts: what it scans, what it produces, and where the result goes. Remaining gaps — no explicit non-destructive statement, no failure/skip semantics for unparseable files, no per-file response structure — are relatively minor given the 'content directly as AI context' framing.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
Schema description coverage is 100% — both patterns and diagramType already have clear descriptions in the schema. The tool description adds only marginal context (e.g., the default 'all workspace files' behavior implied by 'Batch scan and convert all'), so the baseline 3 applies; it neither compensates for gaps nor adds meaningful parameter insight.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
States a specific action ('Batch scan and convert'), specific resources (workspace .xml/.drawio diagram files), and explicit output (Mermaid Markdown). The clause 'returning their content directly as AI context' signals a read-style operation that contrasts with the write sibling, making the purpose unambiguous.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
Usage is implied — this is the tool for getting diagram content into AI context as Mermaid — but there is no explicit when-to-use guidance, no exclusions, and no mention of alternatives such as convert_diagrams_write for persisting output or list_diagrams for merely enumerating files. An agent must infer selection criteria from the name and output framing.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Batch scan workspace diagrams matching glob patterns and convert them to updated .md files on disk.
| Name | Required | Description | Default |
|---|---|---|---|
| patterns | No | Custom glob patterns to scan for diagram files | |
| diagramType | No | Diagram type: "c4" or "uml" (default: "c4") |
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
With no annotations, the description must disclose side effects itself. It does say the tool writes/updates .md files on disk and that it scans in batch, which are important behaviors. However, it doesn't say whether existing files are overwritten, where new files land, or whether the operation is idempotent.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
The description is a single sentence with no filler. The key constraints—batch scanning, glob patterns, and disk output—are front-loaded and every word earns its place.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
For a simple two-parameter tool, the description covers what is scanned and what is produced, but it leaves ambiguity about overwrite behavior and doesn't clarify side effects in detail. Since no annotations or output schema exist, this gap is more significant.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
Schema coverage is 100%, so the baseline is 3. The description adds only 'workspace' and 'batch' context to the patterns parameter; the diagramType enum and default are already in the schema. No additional parameter meaning is needed.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
The description opens with a concrete action ('Batch scan workspace diagrams matching glob patterns') and specifies the outcome ('convert them to updated .md files on disk'). This clearly names the resource and distinguishes it from the sibling tools by emphasizing disk writes.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
It doesn't explicitly state when to prefer this over convert_diagrams_read or list_diagrams. The 'Batch' and 'on disk' wording implies a write-oriented bulk use case, but there are no explicit alternatives or exclusion conditions.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Discover all Draw.io architecture diagram files (.xml, .drawio) in the workspace using glob patterns.
| Name | Required | Description | Default |
|---|---|---|---|
| ignore | No | Custom glob patterns to ignore during search | |
| patterns | No | Custom glob patterns to search for diagram files (e.g. ["**/*.drawio"]) |
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
No annotations are provided, so the description carries the full burden. 'Discover' implies a read-only search operation, and it specifies scope (workspace) and approach (glob patterns). However, it does not explicitly state that the tool returns a list of file paths, does not modify anything, or how defaults behave when no patterns are supplied.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
The description is a single focused sentence that front-loads the purpose, resource type, file extensions, and search method. No unnecessary words or redundant information.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
For a simple listing tool with two optional parameters and full schema coverage, the description is mostly complete. It clearly communicates what files are searched and how. The main gaps are the absence of explicit return-value information and default pattern behavior, but these are minor for this type of tool.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
Schema description coverage is 100%, and both parameters (patterns and ignore) already have meaningful descriptions. The tool description reinforces that glob patterns are involved and that .xml/.drawio files are the targets, but it does not add substantial new meaning beyond the schema.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
The description uses a specific verb ('Discover'), names the resource ('Draw.io architecture diagram files'), and lists the relevant file extensions and method ('glob patterns'). This clearly distinguishes it from sibling tools like convert_diagrams_read and convert_diagrams_write, which are about conversion rather than discovery.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
The description gives clear context: use this when you need to find Draw.io diagram files in the workspace by glob patterns. It does not explicitly mention alternatives or exclusions, but the purpose is distinct enough from the convert siblings that an agent can infer when to choose this tool.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
v1.0.0convert_diagrams_readconvert_diagrams_writelist_diagramsScored across 3 tools
list_diagrams is clearly discovery-only, while the two convert tools are separated by output behavior: _write persists .md files to disk and _read returns content as AI context. There is no meaningful overlap between any pair.
All tool names follow a consistent verb_noun pattern in lowercase snake_case, using the same 'diagrams' resource. The _read/_write suffixes are a coherent way to distinguish output modes, and list_diagrams fits the pattern as an independent discovery action.
Three tools is a well-scoped set for a focused converter utility: discover source files, convert to files, and convert to context. Each tool addresses a distinct workflow step without unnecessary expansion.
The core workflow of finding Draw.io diagrams and converting them to Mermaid Markdown is fully covered in both output modes (disk and context). No obvious dead ends or missing lifecycle operations are apparent for this narrow domain.
Let Claude, Cursor, or ChatGPT author Mermaid diagrams your team can read and share.
Generate dynamic Mermaid diagrams and charts with AI assistance. Customize styles and export diagr…
Create and edit architecture diagrams from your AI agent; get an SVG and a live editable canvas.
AI Agent with Architectural Memory. Impact analysis (free), tests and code from the graph (pro).
