NetMap — Network Diagram MCP
Allows launching PostgreSQL clients (e.g., psql, DBeaver) from the network diagram by clicking on PostgreSQL ports, using URL protocol handlers.
Allows launching Redis clients (e.g., redis-cli, RedisInsight) from the network diagram by clicking on Redis ports, using URL protocol handlers.
@den.dance/network-diagram-mcp
Visual network topology editor with AI agent integration via the Model Context Protocol. Open-source successor to netViz (netViz Inc., 1990 → CA Technologies → discontinued 2012).
Try it live → https://map.den.dance/

What makes it different
One-click ops dashboard. Click any web port (80 / 443 / 8080 / …) on a node card to open the service in a new tab. Click an SSH / Postgres / Redis port → in-app picker offers your installed clients (iTerm / Windows Terminal / psql / DBeaver / redis-cli / RDM / …) and launches the one you choose. The picker re-detects every click — newly installed clients show up immediately, never a silent locked-in default.
AI agents edit the map. 44 MCP tools — your Claude / Cursor / Claude Desktop session can build a diagram for you, search across nodes / ports / notes, run smart auto-layouts, export PNG / PDF — all by talking to the open browser tab.
nmap -oXimport. Run a network scan, hand the XML to the agent, get a map with nodes auto-typed from OS fingerprint + port profile (router / switch / firewall / server / printer / …).CSV inventory import. Drop in your asset spreadsheet — auto-detects columns (name / ip / type / ports / notes; aliases like
hostname/descriptionaccepted).Operator-grade node cards. Per-node: open ports with service names, Docker services, DNS domains, free-form notes. All searchable cross-entity via the agent.
Smart auto-layout. Force-directed (deterministic via seed), cluster-by-type, cluster-by-connection.
Multi-sheet, local-first. Multiple maps in one workspace, everything in
localStorage. No login, no cloud. Export JSON / PNG / PDF.
Related MCP server: Heym
How it works
This package is a stdio → WebSocket bridge. It runs locally as an MCP server; the live NetMap browser tab connects to it over WebSocket. The agent talks to the bridge over MCP; the bridge forwards calls to the browser, which updates React state in real time.
LLM agent (Claude Code / Claude Desktop / Cursor)
│ MCP protocol (stdio)
▼
@den.dance/network-diagram-mcp ← this package, runs locally
│ WebSocket ws://localhost:47821
▼
NetMap in browser (https://map.den.dance/) ← React state updates live⚠️ Requires an open NetMap browser tab. Open https://map.den.dance/ — the MCP bridge is enabled by default (v1.2+), so the connection comes up as soon as the tab is open. Without an open tab the bridge has no peer and every tool call will time out. (To opt out:
Settings → ⚙ → Integrations → AI Agent (MCP) → Enable WebSocket connection.)
Install
Claude Desktop
Add to claude_desktop_config.json:
{
"mcpServers": {
"netmap": {
"command": "npx",
"args": ["@den.dance/network-diagram-mcp"]
}
}
}Claude Code
claude mcp add netmap -- npx @den.dance/network-diagram-mcpWith a custom port:
claude mcp add netmap -e NETMAP_MCP_PORT=12345 -- npx @den.dance/network-diagram-mcpThen open https://map.den.dance/ — the bridge is enabled by default (v1.2+). Settings → Integrations → AI Agent (MCP) if you need to inspect or toggle the connection. The toolbar will show a 🟢 MCP online badge when the browser and server are connected.
Verify the agent side:
claude mcp listYou should see netmap in the list.
Environment variables
Var | Default | Description |
|
| Local WebSocket port the bridge listens on. Must match the URL configured in NetMap's Settings → Integrations → AI Agent (MCP). |
URL Protocol Handlers + in-app client picker (v1.1+, picker since v1.2)
This package also ships an OS-level agent that registers handlers for ssh://,
postgres://, and redis:// URLs. Click any such link in a browser or terminal
and the agent spawns your installed client (psql / redis-cli / iTerm /
Windows Terminal / kitty / DBeaver / TablePlus / RedisInsight / …).
Inside NetMap (v1.2+): when the serve daemon is running, clicking an
SSH / Postgres / Redis port button on a node card opens an in-app picker of
installed clients on your machine and launches the one you pick — every click
re-detects, so newly installed clients show up immediately. Without the daemon,
port-clicks fall back to the OS default URL handler.
Install once per machine:
# Linux: ~/.local/share/applications/netmap-<scheme>-handler.desktop + xdg-mime
# macOS: ~/Library/Application Support/NetMap/handlers/NetMap<Scheme>Handler.app + lsregister
# Windows: HKCU\Software\Classes\<scheme> (per-user, no admin)
npx @den.dance/network-diagram-mcp install --allOther CLI subcommands:
npx @den.dance/network-diagram-mcp install ssh # single scheme
npx @den.dance/network-diagram-mcp uninstall postgres # remove registration
npx @den.dance/network-diagram-mcp list # what's registered (JSON)
npx @den.dance/network-diagram-mcp detect ssh # which clients are available
npx @den.dance/network-diagram-mcp serve # daemon: WS + HTTP on :47821Click handling is daemon-less — OS routes the link to a short-lived
bin/handler.js process which parses the URL, picks the first installed
client (priority order: CLI → popular GUI → cross-platform power), and spawns
it. The optional serve daemon adds HTTP endpoints (/status, /detect,
/exec) for browser-side integration with a Bearer-token-gated /exec.
npx @den.dance/network-diagram-mcp with no args still runs the original
stdio MCP server (back-compat — Smithery / Claude Desktop probes are
unaffected).
Top-5 clients per scheme (priority order, first-installed wins):
Scheme | Clients |
| iTerm2 → Windows Terminal → GNOME Terminal → Terminal.app → kitty |
| psql → TablePlus → DBeaver → Beekeeper Studio → pgAdmin 4 |
| redis-cli → RedisInsight → Another Redis DM → Medis → RDM (legacy) |
postgresql:// aliases to postgres; rediss:// is preserved and triggers
redis-cli --tls.
Available tools
State & layout
Tool | Description |
| Return all nodes, connections, stickies, notes |
| Clear the entire map (requires |
| Naive auto-arrange ( |
| Smart auto-layout: |
Nodes
Tool | Description |
| Add a node ( |
| Update node fields by id |
| Delete a node and its connections |
| Move node to new coordinates |
Connections
Tool | Description |
| Add connection ( |
| Update |
| Delete a connection by id |
Sticky notes
Tool | Description |
| Add a sticky note ( |
| Update |
| Delete a sticky note by id |
| Move sticky to new coordinates |
Sheets (multi-sheet)
Tool | Description |
| List all sheets with metadata and active sheet id |
| Get nodes / connections / stickies for a sheet (default: active) |
| Create a new empty sheet and switch to it |
| Switch active sheet by id |
| Rename a sheet |
| Delete a sheet (requires |
View / canvas
Tool | Description |
| Set zoom level (0.1–3.0) |
| Pan canvas to absolute pixel position |
| Auto-fit all nodes into viewport |
Lock / protection
Tool | Description |
| Lock ( |
Notes & settings
Tool | Description |
| Get sheet-level notes text |
| Set sheet-level notes text |
| Get app settings (sshMode, showGrid, etc.) |
| Update app settings |
Search
Tool | Description |
| Structured node-only filter by |
| Full-text cross-entity search. Looks across nodes (name / ip / notes), stickies (text), connection labels, and ports (port number + service name). Optional |
| Return every node of a single type with FULL field data (ports, dockerServices, domains, ips, notes, parentServer, …). Use this when you need the complete objects, not the stripped projection from |
Example — find everything matching postgres anywhere in the map:
// → call
{ "q": "postgres" }
// → result
{
"nodes": [{ "id": "n1", "name": "db-primary", "type": "server", "ip": "10.0.0.5" }],
"stickies": [{ "id": "s2", "text": "TODO: upgrade postgres 15 → 16" }],
"connections": [{ "id": "c4", "label": "postgres replication", "from": "n1", "to": "n2" }],
"ports": [{ "nodeId": "n1", "nodeName": "db-primary", "port": 5432, "protocol": "tcp", "service": "postgresql" }],
"total": 4
}map_search and map_get_nodes_by_type are read-only — they work on locked sheets.
Import
Tool | Description |
| Replace active-sheet content with a provided |
| Parse |
| Import nodes from CSV. Auto-detects columns from the header row ( |
Example — turn a 2-host nmap scan into a map in one call:
// → call
{
"xml": "<?xml version=\"1.0\"?>\n<nmaprun>\n <host>\n <address addr=\"10.0.0.1\" addrtype=\"ipv4\"/>\n <hostnames><hostname name=\"web.local\" type=\"user\"/></hostnames>\n <ports>\n <port protocol=\"tcp\" portid=\"22\"><state state=\"open\"/><service name=\"ssh\"/></port>\n <port protocol=\"tcp\" portid=\"80\"><state state=\"open\"/><service name=\"http\"/></port>\n </ports>\n </host>\n <host>\n <address addr=\"10.0.0.2\" addrtype=\"ipv4\"/>\n <os><osmatch name=\"Cisco IOS router\" accuracy=\"98\"/></os>\n </host>\n</nmaprun>"
}
// → result
{ "count": 2, "ids": ["…", "…"] }Provide {"hosts": [...]} instead of xml when you already have parsed host data — hosts takes priority when both are given.
Example — turn a CSV inventory snippet into a map:
// → call
{
"csv": "name,ip,type,ports,notes\ndb-primary,10.0.0.5,server,\"22/tcp,5432/tcp\",Postgres 15\nrouter-main,10.0.0.1,router,22/tcp,Edge router"
}
// → result
{ "count": 2, "ids": ["…", "…"] }For non-standard headers, pass an explicit columns mapping (-1 means "absent"):
{
"csv": "Host,Address\nfoo,10.0.0.99",
"columns": { "name": 0, "ip": 1, "type": -1, "ports": -1, "notes": -1 }
}map_import_nmap and map_import_csv are mutations — blocked on locked sheets.
Layout
map_suggest_layout repositions every node according to a chosen algorithm; result shape:
{ ok: true, algorithm, changed: <node count> }.
// Force-directed (organic; reproducible with seed)
{ "algorithm": "force", "iterations": 200, "seed": 42 }
// Group nodes into horizontal lanes by type
{ "algorithm": "cluster-by-type" }
// Place each connected component in its own x-zone
{ "algorithm": "cluster-by-connection" }Force-directed is O(n²) per iteration — fine up to a few hundred nodes; bring iterations down
for larger scenes. map_suggest_layout is a mutation — blocked on locked sheets.
Export (JSON / PNG / PDF)
Tool | Description |
| Export a single sheet as a JSON object (auto-connections included). Legacy entry — still works. |
| Multi-format export: |
// JSON
{ "format": "json" }
// → { format: "json", data: { nodes, connections, stickies, notes, auto_connections } }
// PNG (max 30 s; agent must persist the base64 to a file)
{ "format": "png" }
// → { format: "png", filename: "netmap-<sheetId>.png",
// base64: "iVBORw0KGgoAAAANSUhEUg…",
// mimeType: "image/png" }
// PDF
{ "format": "pdf" }
// → { format: "pdf", filename: "netmap-<sheetId>.pdf",
// base64: "JVBERi0xLjQK…",
// mimeType: "application/pdf" }PNG / PDF capture renders the live workspace via html-to-image + jsPDF, so the MCP server raises
the per-command timeout to 30 s for this tool. map_export is read-only — works on locked sheets.
Connection status indicator
A badge appears next to the NetMap version in the toolbar (click it to open Settings):
Badge | Meaning |
🟢 MCP online | Connected — agent can edit the map |
🟡 MCP | Connecting to server |
🟠 MCP retry N/10 | Retrying, up to 10 attempts × 30 sec |
🔴 MCP error | Gave up — start the server, then toggle off / on to retry |
(no badge) | Disabled in settings |
Detailed status (with URL and hints) is shown inside Settings → Integrations → AI Agent (MCP).
Running from source (for contributors)
git clone https://github.com/den-indance/network-diagram-mcp.git
cd network-diagram-mcp
npm install
node index.jsOverride port: NETMAP_MCP_PORT=12345 node index.js.
Register the local build in Claude Code instead of npm:
claude mcp add netmap-dev -- node /absolute/path/to/network-diagram-mcp/index.jsLicense
MIT — see LICENSE.
Available Tools
44 toolsmap_add_connectionA
Add a manual connection between two nodes. Returns the new connection id.
Inputs: to_id (required), from_id (required), color, label.
map_add_nodeB
Add a new node to the map. Returns the new node id.
Inputs: name (required), type (required), x, y, +1 more.
map_add_stickyA
Add a sticky note to the map. Returns the new sticky id.
Inputs: text (required), x, y, color.
map_arrangeB
Auto-arrange all nodes in a grid or circle layout.
Inputs: layout (required).
map_bulk_add_connectionsA
Add multiple connections in a single call. Returns array of created ids.
Inputs: connections (required).
map_bulk_add_nodesA
Add multiple nodes in a single call. Each node accepts the same fields as map_add_node plus all type-specific fields (ports, dockerServices, domains, parentServer, etc.). Returns array of created ids.
Inputs: nodes (required).
map_clearA
Remove all nodes, connections and stickies from the current sheet.
Inputs: confirm (required).
map_create_sheetA
Create a new empty sheet and switch to it. Returns new sheet id.
Inputs: name.
map_delete_connectionB
Delete a connection by id.
Inputs: id (required).
map_delete_nodeA
Delete a node and all its connections.
Inputs: id (required).
map_delete_sheetA
Delete a sheet. Cannot delete the last remaining sheet.
Inputs: id (required), confirm (required).
map_delete_stickyA
Delete a sticky note by id.
Inputs: id (required).
map_duplicate_nodeA
Clone a node with all its properties (ports, services, domains, etc.). Returns new node id.
Inputs: id (required), name, offset_x, offset_y.
map_exportA
Export current or specified sheet as JSON / PNG / PDF. PNG and PDF return base64-encoded blobs with mimeType — the client decodes and saves to a file. Server raises the response timeout to 30s for PNG/PDF since large maps can take several seconds to render.
Inputs: format, sheet_id.
map_export_sheetA
Export sheet data as JSON including auto-connections. Defaults to active sheet.
Inputs: id.
map_find_nodeA
Search nodes by name, IP, type, or free-text query. Returns id, name, type, ip, x, y for each match. Use this to locate a node before updating it.
Inputs: ip, name, text, type.
map_get_all_connectionsA
Return all connections including auto-generated ones (DNS→server, project→server, service→node). Auto connections have auto:true flag. Use this instead of map_get_state.connections when you need the full topology picture.
Inputs: no named parameters.
map_get_nodes_by_typeA
Return all nodes of a given type with FULL field data (ports, dockerServices, domains, ips, notes, parentServer, etc.). For stripped projection or multi-criteria filtering use map_find_node instead.
Inputs: type (required).
map_get_notesB
Return the sheet-level notes text.
Inputs: no named parameters.
map_get_selectionA
Return currently selected node/sticky ids and connection id. Result shape: { nodeIds: string[], nodeId: string|null (= nodeIds[0], legacy single-id), connId: string|null }.
Inputs: no named parameters.
map_get_settingsA
Return current app settings (sshMode, dblClickAction, perfMode, showGrid, showZoomSlider).
Inputs: no named parameters.
map_get_sheet_dataA
Return nodes/connections/stickies/notes for a specific sheet (defaults to active sheet).
Inputs: id.
map_get_stateA
Return the current map state: all nodes, manual connections, stickies and notes. NOTE: only manual connections are included — use map_get_all_connections to also get auto-generated connections (DNS→server, project→server, service→node).
Inputs: no named parameters.
map_get_viewA
Return current canvas zoom level and pan offset.
Inputs: no named parameters.
map_import_csvA
Import nodes from CSV. Auto-detects columns from the header row (name/ip/type/ports/notes — case-insensitive, aliases like hostname/description/host_ip accepted). Ports field accepts 'port/protocol,port/protocol' (e.g. '22/tcp,80/tcp') or bare port numbers (default tcp). Nodes laid out in a 6-column grid. Returns {count, ids}.
Inputs: csv (required), columns.
map_import_nmapA
Parse nmap scan output (XML from nmap -oX) and create nodes with their open ports. Auto-infers node type from OS fingerprint + port profile (router/switch/firewall/server/printer/etc.). Provide either raw xml OR pre-parsed hosts[] — hosts takes priority when both are given. Nodes are auto-arranged in a grid. Returns {count, ids}.
Inputs: xml, hosts.
map_import_sheetA
Replace active sheet content with provided JSON data. Useful for loading a prepared topology in one call.
Inputs: data (required).
map_list_sheetsA
List all sheets with their id, name, createdAt, updatedAt and which is active.
Inputs: no named parameters.
map_lock_sheetA
Lock or unlock a sheet. A locked sheet rejects all MCP mutations and disables manual editing in the UI. Pass locked: false to unlock.
Inputs: id, locked.
map_move_nodeA
Move a node to new canvas coordinates without changing any other fields.
Inputs: x (required), y (required), id (required).
map_move_stickyA
Move a sticky note to new canvas coordinates without changing its text or color.
Inputs: x (required), y (required), id (required).
map_rename_sheetB
Rename a sheet.
Inputs: id (required), name (required).
map_searchA
Full-text search across nodes (name/ip/notes), stickies, connection labels, and ports (port-number/service). Returns categorized hits with a total count. Use map_find_node for structured node-only filtering.
Inputs: q (required), types.
map_selectA
Programmatically select node(s)/sticky(s) and/or a connection. Pass node_ids (array) for multi-select; node_id (string) is legacy single-select. Empty array or null deselects. Connection selection remains single (multi-select connections out of scope v1).
Inputs: conn_id, node_id, node_ids.
map_set_canvas_offsetA
Pan canvas to an absolute pixel offset.
Inputs: x (required), y (required).
map_set_notesA
Set the sheet-level notes text.
Inputs: notes (required).
map_set_zoomA
Set canvas zoom level (0.1–3.0).
Inputs: zoom (required).
map_suggest_layoutA
Smart auto-layout: force-directed (organic, deterministic with seed), cluster-by-type (horizontal lanes per node type), or cluster-by-connection (BFS components → separate zones). Different from map_arrange (naive grid/circle). Mutates node positions in-place.
Inputs: seed, algorithm, iterations.
map_switch_sheetA
Switch the active sheet. Saves current sheet state first.
Inputs: id (required).
map_update_connectionA
Update label or color of an existing connection.
Inputs: id (required), changes (required).
map_update_nodeA
Update fields of an existing node by id.
FIELD SCHEMA BY NODE TYPE All types: name, ip, x, y, w, h, notes ips: [{address, label}] — extra IP addresses
server / pc / nas / ap / camera / printer / router / switch / firewall / other: ports: [{port, protocol("tcp"|"udp"|"icmp"), service, url, bindIp}] dockerServices: [{name, image, hostPort, containerPort, protocol("tcp"|"udp"), url}]
project: parentServer — id of the parent server node (creates auto-connection) repoUrl — Git/GitLab URL deployPath — deployment directory cicdUrl — CI/CD pipeline URL dockerServices (same as server)
domain (DNS Zone): ip — registrar name (e.g. "Cloudflare") registrarUrl — registrar control panel URL domains: [{name, record("A"|"AAAA"|"CNAME"|"MX"|"TXT"|"NS"|"SRV"), target, url}] target IP in 'target' triggers auto-connection to matching server node
service (External Service): provider — provider name (e.g. "Supabase", "Vercel") dashboardUrl — link to service dashboard serviceConns: [{toId, label}] — links to other nodes (creates auto-connections) servedDomains: [string]
Inputs: id (required), changes (required).
map_update_settingsB
Update one or more app settings.
Inputs: changes (required).
map_update_stickyC
Update text, color or size of a sticky note.
Inputs: id (required), changes (required).
map_zoom_to_fitA
Auto-fit all nodes into the viewport by adjusting zoom and pan.
Inputs: padding.
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
TDQS
Scored across 44 tools
While descriptions are detailed, several tools overlap in purpose: map_get_state, map_get_sheet_data, map_export_sheet, and map_export all return map data; map_arrange and map_suggest_layout both auto-layout; map_find_node, map_search, and map_get_nodes_by_type all query nodes. Explicit notes clarify some boundaries, but an agent could easily choose the wrong tool without careful reading.
All tools share the 'map_' prefix and use snake_case throughout, with most following a verb_noun pattern (map_add_node, map_delete_connection). Minor deviations include bare verbs like map_clear, map_arrange, map_search, and map_export, but these are limited and the overall pattern is predictable.
At 44 tools, the server is significantly over the typical range for an MCP server. While the domain is broad, many tools could be consolidated (e.g., map_get_state vs map_get_sheet_data, map_arrange vs map_suggest_layout) to reduce cognitive load. The high count feels heavy even with the expansive feature set.
The tool surface is remarkably comprehensive, covering full CRUD for nodes, connections, stickies, and sheets, plus view management, settings, search, import (CSV, nmap, JSON), export (JSON/PNG/PDF), layout algorithms, selection, and bulk operations. The domain of network diagram management is well-served with no major dead ends.
Maintenance
Related MCP Connectors
Live, editable canvas for AI agents — author graphs via MCP, then watch and reroute runs live.
Private networking for AI agents: networks, nodes, commands, files, services via hosted MCP.
Deterministic IT asset registry operated by AI agents over MCP. Agents propose, humans approve.
Turn outlines and hierarchical notes into interactive mind maps through a hosted remote MCP server.
Related MCP Servers
- AlicenseNot gradedqualityDmaintenanceCreates and visualizes network diagrams through MCP tools for adding nodes and edges, with real-time web interface and REST API for building architecture diagrams interactively.ISC
- FlicenseCqualityBmaintenanceSelf-hosted, source-available AI workflow automation platform. Build multi-agent, RAG, and tool-using pipelines on a visual canvas and publish any workflow as an MCP server (stdio/SSE/Streamable HTTP). Also an MCP client via the agent node.21,638-
- AlicenseNot gradedqualityFmaintenanceAgent-first knowledge graph MCP server that provides 25 tools for managing a knowledge graph with nodes and edges, plus a human-readable dashboard for LLMs and AI agents.288 npmApache 2.0

Network Sketcherofficial
AlicenseAqualityBmaintenanceLocal-first MCP server that lets AI agents design Cisco-style networks and generate L1/L2/L3 topology diagrams, device tables, and AI-ready context files.1612417Apache 2.0