The premise correction
Miro is cloud-only. There is no on-premises Miro, no self-hosted Miro server, and no enterprise tier that changes that answer. Teams searching this query are therefore not shopping for a Miro edition; they are shopping for an exit, usually driven by one of three forces: data residency rules that forbid workshop content on a vendor's cloud, a security policy that requires software to run inside the company network, or a hard requirement to keep working when the internet connection does not.
That distinction matters because it reframes the comparison. The question is not "which board is most like Miro" but "which board can my team actually run, keep updated, and collaborate on, on our own infrastructure." Graded that way, the field sorts itself quickly, and the sorting key is collaboration: a whiteboard that cannot host a live multi-cursor session on your server is a drawing tool, not a Miro replacement.
The field, graded on operations
| Board | Deployment shape | Live collaboration self-hosted | The operational catch |
|---|---|---|---|
| WBO | One Docker container | Yes, built in | Deliberately minimal feature set |
| Nextcloud Whiteboard | App inside Nextcloud plus a WebSocket collaboration server | Yes, through the separate server component | You are adopting a slice of the Nextcloud stack |
| Excalidraw | Static app plus your own room infrastructure | Not in the stock image; needs a separate room server and custom build work | The easy deployment is a solo canvas |
| CryptPad | Self-hosted CryptPad instance | Yes, with end-to-end encryption | Whiteboard is one app inside a suite; the suite is the commitment |
| AFFiNE | Self-hosted server | Yes | Enterprise backend features sit under a separate source-available license |
WBO (Whitebophir) is the deployment minimalist's answer: a single container, AGPL-3.0, with real-time collaboration as the core feature rather than an add-on. What you give up is everything around the canvas: no template library, no frames, no workshop machinery. For teams whose need is "a shared board on our network, today," that trade is often correct.
Nextcloud Whiteboard is the answer when the organization already runs Nextcloud, or is willing to: the whiteboard is an app in that ecosystem (AGPL-3.0, built on Excalidraw's canvas), and the part to plan for is the separate WebSocket collaboration server that makes live sessions work. Budget the deployment as two components, not one, and test it under real concurrent load before announcing it to the company.
Excalidraw needs the most honest paragraph in this guide. It is the best-known open source canvas (MIT), and its hosted service collaborates beautifully. The self-hosted story is narrower: the official image gives you the drawing application, and live collaboration rooms are a separate piece of infrastructure that you assemble and maintain yourself, historically involving a custom build rather than a switch you flip. Teams routinely discover this after the announcement. If your requirement is a collaborative board on your server this week, Excalidraw self-hosted is a project, not an install.
CryptPad is the encryption-first answer: a full self-hostable office suite (AGPL-3.0 family) whose whiteboard inherits end-to-end encryption, so the server operator cannot read board content. That property is the entire reason to choose it, and it comes with suite-level adoption: accounts, storage quotas, and administration are CryptPad's, not a standalone board's.
What running a board actually costs
Self-hosting trades a subscription for a responsibility list: TLS certificates, updates, backups, uptime during workshops, and user account management. The tools above are all light by server standards; the cost is attention, not hardware. The failure mode to plan for is social, not technical: a board that is down during the quarterly planning session sends the whole company back to a cloud tool on personal accounts within the hour, and the residency problem you solved returns wearing a worse disguise. Assign an owner, automate the backups, and rehearse a restore before the rollout, not after the first incident.
