Updated
When you save in NetworkSketch, you get a .networksketch.json file on your own computer. It's plain JSON: no
encryption, no license check, no account ID inside. Any text editor can open it, any script can read it, and it
goes wherever your other files go: a file share, SharePoint, Git, a ticket attachment.
A saved diagram is readable text. Nothing in it depends on your subscription, so it stays readable whether or not you're a customer.
Diagrams are never uploaded to us. There is no server copy to lose, leak or hold back. How that works.
New versions of NetworkSketch open files saved by older ones. New fields are added alongside the old ones, never in place of them.
The structure below is the real one, so you can generate diagrams from your own inventory or pull data back out with a few lines of code.
Editable PNGs. Every PNG exported from NetworkSketch carries the whole diagram inside it, as a compressed
iTXt chunk with the keyword networksketch holding the same JSON as a saved file. Store the image in Hudu,
IT Glue, Confluence, SharePoint or a ticket, and open it in NetworkSketch (Open, or drag it onto the canvas) to edit the diagram
again. Upload it as a file: an image pasted from the clipboard, resized or re-saved by another program loses the diagram.
The same data also comes out as spreadsheets: Export gives you CSV files of the device inventory, cabling schedule and VLAN/subnet table, alongside the PDF and PNG.
A file is one JSON object. Positions are in pixels from the top left of the page; for anything inside an area, they're relative to that area.
{
"app": "netsketch", // always "netsketch" (files from before the rename too)
"version": 1, // format version
"doc": { ...title block... },
"pages": [ { "id", "name", "nodes": [...], "edges": [...] } ],
"activePageId": "p_k3x9a1bc" // the page that was showing when it was saved
}
| Key | What it holds |
|---|---|
title, customer, address | Diagram title, customer or site, and its address |
author, approver, approvalDate | Who drew it and who approved it |
version, changedAt, lastEditor | Revision number, date of the last change (YYYY-MM-DD) and who made it |
verifiedBy, verifiedOn | Who last checked the diagram against the real network, and when |
changeLog | A list of { "date", "by", "note" } entries, printed on the PDF's change log page |
Every node has an id, a type, a position { "x", "y" } and data. Areas and notes also have style { "width", "height" }, and anything inside an area has parentId, the area's id.
| type | data |
|---|---|
device | kind (see below), label (the hostname), labelPos, and props with the device's fields |
area | kind is site, room, vlan, zone, vpc or rack; props holds cidr, vlanId, gateway, dhcp, location and so on |
note, text, legend | Annotations: props.text for notes and labels; the legend is the Media Type Key |
Device fields in props (each device type uses the ones that apply): mgmtIp (one or more IPv4/IPv6 addresses, comma-separated), mac, vendor, model, serial, assetTag, os, role, location, ports, vlans, ha, notes, plus type-specific ones such as circuitId, carrier, bandwidth and publicIps on circuits, ssids on access points, or dids and channels on SIP trunks. All values are strings.
Device kinds: isp internet mpls cellular cloudprov router sdwan lte modem loadbal switch l3switch stack patch firewall vpn nac ap wlc bridge server hypervisor vm storage database dirsvc workstation laptop printer mobile camera endpoint ipphone pbx sbc voicegw ups pdu siptrunk ccaas, and goto for a link to another page (props.page is the target page's id).
Each edge has an id, source and target (node ids), type "link", and data:
| Key | What it holds |
|---|---|
media | copper, smf (single-mode OS1/OS2), mmf12 (OM1/OM2), mmf34 (OM3/OM4), mmf5 (OM5), dac, wireless, wan, vpn, sip or logical |
srcPort, dstPort | The interface at each end, e.g. Gi1/0/48 |
speed, vlans, label, notes | Link speed, VLANs carried, a label on the line, free notes |
cableId, length, test | Cable label, length and certification test result, for the cabling schedule |
routing | How the line is drawn, e.g. orthogonal |
sourceHandle and targetHandle only record where a drag started; a cable is undirected, so source and target are interchangeable.
A firewall and a switch, the switch in a VLAN area, joined by a fiber uplink. Save this as example.networksketch.json and open it in the editor:
{
"app": "netsketch",
"version": 1,
"doc": { "title": "Branch 12", "customer": "Acme", "version": "1.0", "changedAt": "2026-10-24" },
"pages": [{
"id": "p_main", "name": "Physical",
"nodes": [
{ "id": "n_users", "type": "area", "position": { "x": 0, "y": 160 }, "style": { "width": 320, "height": 160 },
"data": { "kind": "vlan", "label": "Users", "props": { "vlanId": "20", "cidr": "10.12.20.0/24", "gateway": "10.12.20.1" } } },
{ "id": "n_fw", "type": "device", "position": { "x": 120, "y": 0 },
"data": { "kind": "firewall", "label": "BR12-FW01", "labelPos": "right", "props": { "mgmtIp": "10.12.0.1", "model": "FortiGate 60F" } } },
{ "id": "n_sw", "type": "device", "position": { "x": 120, "y": 60 }, "parentId": "n_users",
"data": { "kind": "switch", "label": "BR12-SW01", "labelPos": "right", "props": { "mgmtIp": "10.12.20.2" } } }
],
"edges": [
{ "id": "e_up", "source": "n_fw", "target": "n_sw", "type": "link",
"data": { "media": "mmf34", "speed": "10G", "srcPort": "port1", "dstPort": "Te1/1/1", "vlans": "20" } }
]
}],
"activePageId": "p_main"
}
Areas come before the devices inside them in a page's nodes list.
Yes. Write the JSON above from a script (a CMDB export, a spreadsheet, NetBox, a CDP/LLDP dump) and open the file in NetworkSketch. Positions can be rough: Auto-arrange lays the page out for you. For a spreadsheet, Import reads a device list directly.
Yes, if it's the PNG NetworkSketch exported (or a copy of that file). The diagram is stored inside the image, so Open or dragging it onto the canvas brings back every page, device, port and cable. Documentation tools that keep uploaded files as they are (attachments, file uploads) keep the diagram; pasting the image from the clipboard or a tool that resizes uploads removes it, and NetworkSketch tells you when an image has no diagram inside.
Yes. Files are saved indented, one key per line, so diffs between revisions are readable.
No. CallSketch call flows have their own structure, and each editor tells you if you open the other one's file.
Autosave keeps the open diagram in your browser's local storage so a closed tab doesn't lose work. It's the same JSON, on your computer only. Saving to a file is what makes it permanent.