cisco-open/network-sketcher

Network Sketcher is an AI-ready network design tool with Local MCP, Online, and Offline editions for creating network designs and exporting PowerPoint diagrams and Excel-based configuration data.

378 ★23 forksPythonUpdated 3d ago

What you need to know

AI-native network design and diagramming tool that generates L1/L2/L3 Cisco-style topology diagrams, device tables, and AI-ready context files from structured network data; the Local MCP edition exposes it as an MCP server so LLM clients like Cursor and Claude Code drive network design via tool calls.

Install

git clone https://github.com/cisco-open/network-sketcher/
cd network-sketcher/network-sketcher_local_mcp && python -m pip install -r requirements_mcp.txt

Usage

  • Add to Cursor mcp.json: { "mcpServers": { "network-sketcher": { "command": "python", "args": ["/path/to/network-sketcher/network-sketcher_local_mcp/ns_mcp_server.py"] } } }
  • Or Claude Code: claude mcp add network-sketcher -- python "/path/to/.../ns_mcp_server.py"
  • Ask e.g. "create a small 5-site WAN design with HQ, two data centers, two branches... then build the default L1/L2/L3 diagrams and device table"
  • Platform import workflow: run a Cisco Extension converter, feed ns_commands.txt through run_commands, then finalize with build_default_outputs

Key features

  • Generates L1 physical, L2 VLAN/broadcast-domain, and L3 IP topology diagrams (SVG/PowerPoint) plus HTML device tables
  • AI Context file export so another LLM can review or extend the network design
  • Three editions: Local MCP (AI-native, stdio), Online (browser web service), Offline (desktop GUI + CLI)
  • Network Sketcher Cisco Extension converters import ACI, Catalyst Center, Meraki, Nexus Dashboard/NDFC, CML, Cyber Vision, SNA, NetBox exports, and config files (IOS/IOS-XE/NX-OS/IOS-XR/ASA)
  • Offline edition supports conversion from Visio, draw.io, NetBox, and CML data; communication flow management

Best for

AI-driven Cisco-style network design and topology diagram generation for network engineers using Cursor or Claude Code.

Caveats

  • Local MCP: single-user, stdio transport only (no HTTP/SSE), requires Python 3.10+ and the full repo (network-sketcher_online folder must be present)
  • End-to-end verified in Cursor (primary tested host); Claude Code support is design-validated only at this release
  • Recommended LLM: Claude Opus 4.7 or later — weaker models may struggle with the multi-step tool workflows
  • LLM clients cannot directly view binary PPTX/SVG output; diagram generation for large networks may take time
  • Online/Offline editions are IPv4 only (no IPv6); Online edition is tested on Windows only and intended for internal networks
Platforms: Windows · macOS · LinuxClients: Cursor · Claude Code
Documentation ↗

Reviewed 2026-08-07

Topics

ai-agentciscoclaude-codecursormcp-servermodel-context-protocolnetwork-automationnetwork-diagrampowerpointsvgtopology
View on GitHub ↗
Stars
378★
Forks
23
Language
Python
License
Apache-2.0
Created
2023-08-08
Last push
2026-08-04