Network Sketcher (Local MCP)

What is Network Sketcher (Local MCP)?

Network Sketcher (Local MCP) is an AI-native network design and diagramming tool that exposes the Network Sketcher engine as a Model Context Protocol (MCP) server. It lets LLM clients such as Cursor and Claude Code design Cisco-style networks directly through tool calls, running locally over stdio and keeping master files on the user’s machine.

How to use Network Sketcher (Local MCP)?

Clone the repository, install the Python dependencies (Python 3.10+ required), then add the server to your MCP client’s configuration (e.g., Cursor’s mcp.json or Claude Code’s claude mcp add command). After setup, ask the agent to build a network using natural-language prompts; the generated SVG diagrams, PowerPoint files, and device tables are saved in the workspace.

Key features of Network Sketcher (Local MCP)

• No browser or copy-paste needed; LLM drives design via Tool calls
• Generates L1/L2/L3 topology diagrams, device tables, and AI context files
• Reuses the existing network-sketcher_online/ns_engine/ as a library
• Supports importing from live inventory via the Cisco Extension converters
• Single-user edition designed for local PC operation over stdio transport

Use cases of Network Sketcher (Local MCP)

• Build a 5-site WAN or campus LAN design from an AI conversation
• Generate L1 physical, L2 VLAN/broadcast-domain, and L3 IP topology diagrams
• Export a combined L1/L2/L3 HTML viewer and an interactive device table
• Create an AI Context file so another LLM can review or extend the network design
• Import from live inventory (ACI, Catalyst Center, Meraki, etc.) via separate converters

FAQ from Network Sketcher (Local MCP)

How does Network Sketcher (Local MCP) differ from the Online and Offline editions?

The Local MCP edition is AI‑native: the LLM executes CLI commands directly via MCP Tool calls without a browser or copy‑paste. The Online edition uses a browser with human‑LLM copy‑paste, and the Offline edition is a desktop GUI/CLI without native AI integration.

What are the runtime requirements?

Python 3.10 or later, the full Network Sketcher repository (including the network-sketcher_online/ folder), and an MCP‑compatible client. A model like Claude Opus 4.7 or later is recommended due to the multi-step workflow.

What transport does Network Sketcher (Local MCP) support?

Only stdio transport is supported. HTTP/SSE transport is not available.

What are the main limitations?

Single‑user local operation; diagram generation may be slow for large networks; binary output (PPTX/SVG) cannot be directly viewed by the LLM client; end‑to‑end verification is primarily done in Cursor (Claude Code is design‑validated but may have issues).

Does Network Sketcher (Local MCP) store any data externally?

No. All master files are kept on the user’s local machine. The server communicates only via stdio with the LLM client.