excalidash-ai-mcp-architecture

Generates MCP server architecture diagrams with explicit transport, auth, registry, and storage boundaries.

2|Updated Jun 17, 2026
One-click install
npx skills add https://github.com/gabedsam01/excalidash-v2 --skill excalidash-ai-mcp-architecture-gabedsam01
Or copy as Structured Prompt for Agentβ–Ό
Please help me install this Agent Skill.
Skill: excalidash-ai-mcp-architecture
Source: https://github.com/gabedsam01/excalidash-v2/tree/main/skills/excalidash/excalidash-ai-mcp-architecture
Command: npx skills add https://github.com/gabedsam01/excalidash-v2 --skill excalidash-ai-mcp-architecture-gabedsam01

SYSTEM DOCUMENTATION & REQUIREMENTS

πŸ’‘ This Skill includes scripts (resource) and references (resource) components.

What problem does it solve? Drawing an accurate MCP server or AI tool architecture diagram is error-prone: the LLM often ends up inside the server trust boundary, auth gets collapsed into the transport node, and bearer keys or database URLs leak into the drawing. This Skill produces a validated Excalidraw architecture diagram with the correct lane separation and zero leaked secrets. ## Core Features & Use Cases - Structured MCP architecture generation: Builds four lanes β€” client (host + LLM, outside the trust boundary), transport (/mcp JSON-RPC), server (bearer auth -> tool registry -> services), and backend (PostgreSQL + library cache) β€” via the ExcaliDash MCP tools. - Quality-gated workflow: Runs a lint -> score -> repair loop until the diagram scores at least 95 with zero hard blockers, then validates architecture separation before saving and exporting. - Secret redaction: Redacts bearer tokens, database URLs, and JWT secrets before they reach any tool argument, and re-scans exports as a backstop. - Use Case: Ask to diagram your ExcaliDash MCP server and receive a saved, shareable SVG/PNG showing the host and LLM outside the trust boundary, the exd_ bearer gate as an explicit node, and storage reachable only through services. ## Quick Start Ask the AI to draw the architecture of your MCP server with transport, bearer auth, the tool registry, and PostgreSQL storage as separate nodes, keeping the LLM outside the trust boundary.

Frequently Asked Questions about excalidash-ai-mcp-architecture

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I draw an MCP server architecture diagram with Excalidraw?β–Ό

Describe your server β€” host, /mcp JSON-RPC transport, bearer auth, tool registry, services, and PostgreSQL storage β€” and the skill generates the diagram via apply_architecture_skill with pattern mcp. It then lints, scores, and repairs the drawing until it reaches a score of at least 95.

How do I show where the LLM sits relative to my MCP server?β–Ό

The skill places the host and the LLM in a client lane outside the server trust-boundary frame, with the LLM visually distinct. The validate_architecture tool confirms both are outside the boundary before the diagram is saved.

Can I generate an architecture diagram from a repository analysis?β–Ό

Yes. Pass a repository analysis object with modules, entrypoints, database, services, and integrations to create_from_repo_analysis, and it maps those elements onto the client, transport, server, and backend lanes automatically.

When should I not use the MCP architecture pattern?β–Ό

Do not use it for LLM or RAG data pipelines, generic microservices without an MCP transport and tool registry, C4 container views, deployment topology, or full threat models. Those shapes belong to the dedicated pipeline, microservices, C4, DevOps, or security skills.

How are API keys and database URLs handled in generated diagrams?β–Ό

Secrets are redacted before reaching any tool argument: bearer keys become Bearer exd_[REDACTED_API_KEY] and Postgres URLs become postgres://exd:[REDACTED_DATABASE_URL]@db:5432/excalidash. Exports are re-scanned as a backstop so no raw credential appears in the drawing.