The 10 Best Open Source Miro Alternatives in 2026

Key Takeaways

• People ask for open source for four reasons: cost, self-hosting, data control, and no vendor lock-in. Only two are decided by the licence.
• Excalidraw ranks first on an MIT licence, a real maintenance record, and the shortest distance between wanting a canvas and having one.
• Self-hosted Excalidraw has no server-side persistence out of the box, which surprises everyone who deploys it expecting a Miro replacement.
• tldraw v2 and later ship under a source-available licence, not an OSI-approved one. Removing the watermark commercially costs money.
• AGPL v3 is the licence that bites here, because a whiteboard is a network service. CryptPad, Nextcloud, WBO and Logseq all use it.
• Open-Sankoré is dead and Focalboard's standalone product was archived, yet both still appear on current lists. Read the commit log, not the star count.
• Storyflow is closed source and is not ranked here. It appears only in a labelled closing note for readers whose real requirement was price.

Four Reasons, Two Columns

Nobody wants an open source whiteboard for its own sake. You are asking for one of four things, usually without separating them: you want it free, you want to run it on your own machine, you want your data somewhere you control, and you want to be sure nobody can raise the price 40 percent or sell the company to someone who deprecates it.

Those four sort into two columns, and the whole decision comes from noticing which is which.

The licence column holds what the LICENSE file gives you, and it contains exactly two things. Permission: the right to build, run and modify the code without asking anyone. And exit: the right to fork the project if the maintainers stop, relicense, get acquired, or go somewhere you hate.

The deployment column holds what your choice of where to run it gives you. Cost sits here. Data custody sits here. Sign up for Excalidraw Plus and your boards live on their servers exactly as a Miro board lives on Miro's, and the MIT licence underneath has not moved a single byte closer to you.

That is the trap. The licence gives you permission and an exit. Everything else you still have to run yourself.

If cost is your reason, the licence is a poor proxy: a VPS with backups, TLS renewal and a Postgres you now own is not free, it is billed differently. If data control is your reason, ask about deployment first, because a self-hostable closed source tool puts your bytes on your disk while an open source tool on the vendor's cloud does not.

What the licences mean in practice

AGPL v3 is the one to read twice, because a whiteboard is a network service by definition. Deploy it unmodified and you have nothing to publish. Patch it, expose it to users, and you owe them the source of your version.

At a Glance: The 10 Tools Compared

How I Ranked These

I come from documentary, where the archive outlives the production company, and I have watched projects lose their planning materials because a tool was retired. That is what made the exit question real rather than theoretical. I deployed each self-hostable tool on a small VPS and read the LICENSE file in each repository rather than the marketing page.

1. What does the licence actually say? The test: open the LICENSE file at the tag you intend to deploy. Two projects here use language a careful reader would not call open source.

2. What does self-hosting cost every month? The test: count the services in the docker-compose file. One is a weekend. Five is a name on the rota.

3. Does it remember anything? The test: draw something, close the tab, reopen the URL in another browser. Several well loved canvases fail this once self-hosted.

4. Is it genuinely maintained? The test: read the last 90 days of commits. All dependency bumps means it is being kept alive rather than developed.

5. Is there a funded path? The test: does a company or institution pay someone. Unfunded projects stop the week the maintainer changes jobs.

Pricing is as of August 2026 and changes frequently. Verify with each vendor.

Quick Picks by Job Type

• Best overall: Excalidraw. MIT, maintained, instantly usable.
• Best docs and canvas in one: AFFiNE. Self-hosted Notion plus Miro, roughly.
• Best structured diagrams: Draw.io. Apache 2.0, unbeatable shape libraries.
• Best data custody: CryptPad. Encrypted before it leaves the browser.
• Best five minute self-host: WBO. One process, no database, no configuration.
• Best if you run Nextcloud: Nextcloud Whiteboard. No new stack, no new logins.
• Best to build a product on: tldraw. Read the licence terms first.
• Best when the canvas work is design work: Penpot. A serious Figma alternative.

1. Excalidraw

The verdict. The default answer, and the only tool here strong in both columns.

Best for. Sketching and quick sessions where the board is a conversation, not an archive.

Pricing. MIT and free to self-host. Excalidraw Plus starts around $6 per user monthly as of August 2026.

Why it ranks here. MIT lets you fork it, close your fork, ship it commercially, and owe nobody anything beyond a copyright notice. No patent clause, no network use trigger, no watermark, no revenue threshold.

Self-hosting gives you a static frontend with no server-side persistence: boards live in browser local storage. Collaboration needs excalidraw-room, which relays rather than stores. Teams deploy it, invite four people, close the tab, and find the shared workspace was only ever on their own laptop.

That is a design decision, not a defect, and it shares a root with the end to end encryption on excalidraw.com. Persistent boards with accounts are what Plus sells.

Strengths.

• MIT, no conditions beyond attribution.
• Real release cadence and a funded commercial arm.
• No signup, so teams adopt it easily.
• Hand-drawn rendering keeps rough work looking rough.

Limitations.

• No server-side persistence or accounts when self-hosted.
• Collaboration rooms relay rather than archive.
• No voting, timers, private mode or AI.

The trade off. The cleanest licence and the shortest path to a canvas, if you accept it is a drawing surface first.

2. AFFiNE

The verdict. The only project here replacing a docs tool and a whiteboard together.

Best for. Teams wanting an infinite canvas and long-form documents in one self-hosted app.

Pricing. Free to self-host. AFFiNE Cloud has a free tier, with paid plans from roughly $7 monthly as of August 2026.

Why it ranks here. A page and a canvas are two views of the same content. Write in page mode, switch to edgeless mode, and the blocks become objects on an infinite canvas.

On the licence column, check for yourself. AFFiNE's licensing has changed during the project's life and different parts of the codebase have carried different terms. It is open source and the community edition is real, but "read the LICENSE file at the tag you deploy" is specific advice here, not boilerplate.

The deployment column is heavy: Postgres and Redis mean a database you now back up and upgrade. Pin a version and snapshot before migrating.

Strengths.

• Documents and canvas are the same content in two views.
• Local-first, so it works offline before syncing.
• Genuinely open source community edition, actively developed.
• Real-time collaboration self-hosted, not only in the cloud.

Limitations.

• Licence terms have changed. Verify at your deployment tag.
• Postgres and Redis make this a stack, not a container.
• Fast releases mean migrations, and no facilitation features.

The trade off. The most capable open source answer, in exchange for the most infrastructure to own.

3. tldraw

The verdict. The best canvas engine in existence, and not open source in the sense you mean.

Best for. Developers building a canvas into their own product, with a licence budget.

Pricing. Free with the tldraw watermark. Removing it commercially requires a paid business licence as of August 2026.

Why it ranks here. This is the entry most lists get wrong. tldraw v2 and later ship under tldraw's own source-available licence rather than an OSI-approved one. An article calling tldraw MIT is describing a version from years ago.

That is a legitimate business model and tldraw has been open about it. It is still a hard no if your reason was the licence column: source-available gives partial permission and much weaker exit, because you cannot fork and continue on your own terms.

What you get instead is the best canvas engine anyone has shipped. Rendering, selection, arrow binding and sync are a class above everything else here.

Strengths.

• The best canvas rendering and interaction model on this page.
• A mature multiplayer sync layer rather than a bolt-on.
• Excellent extension API for custom shapes and tools.
• Commercially funded, so maintenance answers itself.

Limitations.

• Not OSI-approved. Source-available with commercial conditions.
• Watermark on free commercial use, removable only by paying.
• No unconditional fork right, and hosting is your job.

The trade off. Take it if you are building. Skip it if you need a licence with no conditions attached.

4. Draw.io (diagrams.net)

The verdict. Apache 2.0, old in the good sense, still the best structured diagramming tool you can self-host.

Best for. Architecture diagrams, flowcharts and network maps, where shapes carry meaning.

Pricing. Free, and free to self-host. The Confluence and Jira plugins are commercial as of August 2026.

Why it ranks here. Apache 2.0 is MIT plus two useful things: an explicit patent grant, which corporate legal teams care about, and a requirement to state your changes. If procurement asks about patent exposure, that grant beats MIT.

Deployment is the easiest here after WBO. A single container serves a client-side app that stores nothing, and files live wherever you point it. There is no database to migrate because there is no database, which is why it has outlived two decades of tooling churn.

Where it is not a Miro alternative is collaboration. It is a diagram editor, not a live shared canvas, and free association feels wrong because everything connects to something.

Strengths.

• Apache 2.0 with an explicit patent grant.
• Stateless single container, so maintenance is minimal.
• Enormous shape library covering AWS, Azure, UML and BPMN.
• Desktop app for offline work, with editable XML in exports.

Limitations.

• No real-time multiplayer editing when self-hosted.
• Not a brainstorming surface. Sticky-note work feels wrong.
• Dated interface, and no AI features of any kind.

The trade off. If your Miro boards are mostly boxes and arrows, this replaces them and costs almost nothing to run.

5. Nextcloud Whiteboard

The verdict. The right answer for exactly one audience: teams already running Nextcloud.

Best for. Organisations that already self-host and want the whiteboard inside it.

Pricing. Free and AGPL v3. Nextcloud Enterprise support is quoted per user as of August 2026.

Why it ranks here. If you already run Nextcloud you own the hard parts: users, groups, permissions, sharing, backups and TLS. The whiteboard becomes a file type that inherits your access control instead of introducing a new one.

The wrinkle illustrates the framework neatly. The app is AGPL, and the editor inside it is tldraw, used under an arrangement between the two projects. An AGPL wrapper around a source-available engine is worth understanding before your legal review finds it.

It also needs a separate websocket backend for real-time collaboration, a common source of misconfiguration behind reverse proxies. Budget an afternoon for proxy headers.

Strengths.

• Inherits Nextcloud's users, groups and sharing with no new accounts.
• Boards are files, covered by your existing backups.
• AGPL v3, so the exit guarantee on the app is strong.
• Good canvas quality, because the editor is tldraw.

Limitations.

• The websocket backend is the usual reverse-proxy headache.
• Pointless unless you already run Nextcloud.
• Follows Nextcloud's release train, with no facilitation controls.

The trade off. Nearly free if you are already a Nextcloud shop, and a poor reason to become one.

6. CryptPad

The verdict. The only tool here where the server genuinely cannot read your board.

Best for. Anyone whose real reason was data custody rather than the licence.

Pricing. Free to self-host, AGPL v3. Paid hosted tiers start at a few euros monthly as of August 2026.

Why it ranks here. CryptPad encrypts content in the browser before it reaches the server, so the operator holds ciphertext and key material never leaves the client. That is stronger than "we do not look at your data", because it is an architecture rather than a promise.

It sits sixth because the whiteboard itself is modest. CryptPad's strength is the suite: documents, spreadsheets, kanban, forms and a whiteboard under one encryption model. As a canvas alone it is lighter than Excalidraw or AFFiNE.

Self-hosting is the fiddliest here. It needs a second sandbox domain plus specific HTTP headers, and getting that wrong weakens the isolation the design depends on.

Strengths.

• End to end encryption by architecture, not by policy.
• AGPL v3 with an institutional backer rather than a lone maintainer.
• A whole suite: documents, sheets, kanban, forms, whiteboard.
• Public instances let you evaluate before deploying.

Limitations.

• The whiteboard is basic next to Excalidraw or AFFiNE.
• Needs a second sandbox domain and precise headers.
• Encryption rules out server-side search across your content.

The trade off. Best in class custody in exchange for the least capable canvas in the top half.

7. WBO (WhiteBophir)

The verdict. The lowest effort self-host on this page, and it does one thing properly.

Best for. Getting a private real-time whiteboard on your own server before lunch.

Pricing. Free, AGPL v3. A public instance exists for testing as of August 2026.

Why it ranks here. Count the services: one. WBO is a Node application with no database, storing boards as files on disk. No Postgres, no Redis, no migration that can eat your data, because the data is just files.

For the reader whose deployment column matters most, that simplicity is the feature. A tool a non-specialist keeps running for three years unattended beats a richer tool that rots when the person who set it up leaves.

The scope is honestly narrow: pens, shapes, text, images, live cursors, one board per URL. No accounts, no comments, no templates.

Strengths.

• One process, no database, files on disk.
• Real-time multi-user drawing with no extra service.
• A small, readable codebase you could actually audit.
• Backups are a cron job and a tarball.

Limitations.

• No accounts. Access control is an unguessable URL.
• No comments, templates, frames or structure.
• A very small maintainer pool, so the bus factor is real.

The trade off. The least capable tool in the top seven and the one most likely to still be running in 2029.

8. Penpot

The verdict. Not really a whiteboard, and the best answer if your canvas work is design work.

Best for. Teams replacing Figma rather than replacing Miro.

Pricing. Free and MPL 2.0. Penpot's hosted cloud is free with paid tiers as of August 2026.

Why it ranks here. MPL 2.0 is the useful middle ground: file-level copyleft. Changes to Penpot's own files must stay MPL and be published, but you can combine it with proprietary code without that code becoming MPL.

A large share of the boards people keep in Miro are really UI work: flows, wireframes, component states. Penpot does those properly, with real design primitives, components, prototyping and developer handoff, all of which a whiteboard fakes at best.

It ranks eighth because as a freeform ideation canvas it is worse than Excalidraw. The tool has opinions about design, and they get in the way of scribbling.

Strengths.

• MPL 2.0, genuine open source with a workable commercial path.
• Real design primitives and components, not approximations.
• The self-hosted build matches the cloud version.
• Developer handoff with inspectable CSS and SVG.

Limitations.

• Not a brainstorming surface. Sticky notes are awkward.
• Several services to run, so a real deployment to maintain.
• Heavier in the browser, with a learning curve for non-designers.

The trade off. The right tool for the subset of your Miro boards that were secretly design files.

9. Logseq Whiteboards

The verdict. A good idea attached to a project that spent two years rebuilding its foundations.

Best for. Existing Logseq users who want to arrange their own notes spatially.

Pricing. Free, AGPL v3. Logseq Sync is a paid optional service as of August 2026.

Why it ranks here. The objects on the board are your actual notes, so a block you drag onto the canvas is the same block that lives in your graph rather than a copy. That solves the problem every whiteboard has, which is becoming a dead snapshot of thinking that has moved on.

The honest caveat is maintenance. Whiteboards arrived in 2023, and the team then committed to a rewrite onto a database backend that has absorbed most of the project's attention since. As of my last check in mid 2026 the whiteboard had seen limited feature work. Read the current release notes rather than assuming progress.

There is nothing to self-host: Logseq is a local desktop app over markdown files you own.

Strengths.

• Board objects are live references to your notes, not copies.
• Markdown files in a local folder, so custody is total.
• AGPL v3 with an active overall project.
• Works completely offline with no server.

Limitations.

• No real-time collaboration. This is a single-player tool.
• Whiteboard development slowed during the database rewrite.
• Only useful if you already use Logseq.

The trade off. The best thinking tool here for one person, and no use at all for a team session.

10. OpenBoard

The verdict. A classroom whiteboard for a projector, honestly maintained, and not a Miro alternative.

Best for. Teachers with an interactive display or a graphics tablet.

Pricing. Free, GPL v3, as of August 2026.

Why it ranks here. OpenBoard is included because it keeps appearing on open source whiteboard lists and readers deserve to know what it is. It is a desktop application for teaching: annotate over slides, draw on a projected screen, record the lesson, work page by page.

It grew out of Open-Sankoré and Uniboard and is maintained by a Swiss educational consortium. It is good at that job and completely wrong for a distributed team. No cloud, no accounts, no collaboration. One person, one machine, one screen.

Maintenance is slow but real, driven by institutional need rather than a roadmap, and institutionally funded software often outlives venture funded software by a decade.

Strengths.

• GPL v3 with institutional backing rather than one maintainer.
• Purpose built for teaching over projected content.
• Runs entirely offline on Windows, macOS and Linux.
• Lesson recording and podcast export built in.

Limitations.

• No collaboration. Single user, single machine.
• Page based, not an infinite canvas.
• Slow release cadence and a dated interface.

The trade off. Excellent for a classroom, irrelevant for a remote team, and listed so you find that out here.

What to Actually Pay For

Pay for hosting, not for the licence. Excalidraw Plus at roughly $6 per user and AFFiNE Cloud at roughly $7 monthly fund the open source work you benefit from and remove the operational burden, and you keep the exit right because the code underneath is still open.

Pay a person if you self-host anything with a database. AFFiNE, Penpot and CryptPad are not weekend projects. If nobody's job description mentions them, the deployment drifts and becomes a security problem.

Pay tldraw if you are building a product. The business licence costs less than six months spent rebuilding a canvas engine badly.

Do not self-host to get "free". A $5 VPS is $60 a year plus your time. Self-host for custody and exit, not to save money on a two person team.

Tools to Avoid for This Job

Anything described as open source that ships a source-available licence. BUSL, SSPL, Elastic v2 and bespoke vendor licences are legitimate business models and are not open source. A list calling one of them open source has not read the LICENSE file.

Open-Sankoré. The dead predecessor OpenBoard forked from, still listed as current. If you want that tool, install OpenBoard.

Focalboard as a standalone product. Mattermost stopped developing it separately, and adopting the archived build in 2026 is a dead end.

A repository whose last real feature commit is two years old. Translation merges and dependency bumps are not development.

A self-hosted whiteboard with no backup plan. The commonest way to lose boards is a VPS nobody snapshotted. Lock-in is a slow risk. An unbacked-up Postgres is a fast one.

A Closing Note on Storyflow (Not Open Source)

Storyflow is our product. It is closed source, it cannot be self-hosted, and it has no business being ranked on this list. Everything above stands without it, and Excalidraw is the pick.

This note is for one reader: the person who typed "open source Miro alternative" because what they actually wanted was column two. If your reason was cost, or predictable pricing, or escaping per-seat billing, the licence was never the thing you were shopping for.

Storyflow is an AI canvas for planning creative work. Pricing is flat per account rather than per seat, so inviting people does not raise the bill, and anyone a paid member invites signs up free. Plus is $7.99 monthly billed annually or $9.99 monthly, adding 200 plus Story blueprints and unlimited file uploads. Pro is $14 or $19 monthly, adding AI image generation, roughly twenty times more AI usage and memory across conversations. Max is $39 or $49 monthly, adding forty times more AI and Team Workspace with permissions and roles. It is paid only during early access; the Free plan lands before the end of 2026.

What it does not do: no source code to read, no self-hosted or on-premise deployment, no data residency choice, no fork right of any kind. If it shut down you would have exports and nothing else. It also has no dot voting, no session timers and no facilitator private mode. If your requirement is genuinely a licence, go install Excalidraw. The licence gives you permission and an exit. Everything else you still have to run yourself.

What No Tool on This List Does

None of them replaces Miro's facilitation layer. Dot voting, timers, private mode and breakout frames are what Miro sells to facilitators, and no open source project here has built them.

None has the template library. Miro ships thousands of ready-made boards, and open source projects lack the content operation to match it.

None, including Storyflow, gives you enterprise compliance out of the box. SOC 2, ISO 27001, SSO with SAML, SCIM and audit logs are absent from every tool here. Self-hosting can satisfy a security review another way, but it does not hand you a certificate.

And no licence protects you from yourself. Open source gives you the right to keep running the software. It does not give you the disk it was running on.

The Bottom Line

Install Excalidraw. For nine readers in ten it is the answer: MIT licensed with no conditions worth reading twice, maintained at a real pace, and usable before you finish deciding. If you need something it does not do, the second choice follows the job rather than the licence: AFFiNE for documents plus canvas, Draw.io for structured diagrams, CryptPad for custody, WBO if the deployment has to be maintenance-free.

Then read the LICENSE file yourself. Two of the best known names in this category are not open source in the sense most people mean, and the lists saying otherwise are copying each other rather than the repository.

And be honest about which column you were shopping in. If you came for the fork right and your own hardware, open source is the only thing that delivers those. If you came because you wanted it cheap or wanted out of per-seat billing, you searched the wrong term. The licence gives you permission and an exit. Everything else you still have to run yourself.