Post-fire recovery
Mount Sentinel-2 composites before and after the burn. Draw sectors, annotate regrowth density, run vegetation-index queries and export a PDF report the agency can read.
Public beta — soon
Driades is a collaborative geospatial canvas that runs entirely in your browser. Drop cloud-native data, draw with your team in real time, scrub through years of satellite imagery — all on the same map, with zero installs and zero servers to manage.
Open source and self-hostable from day one. Built for the h4ck1ng.science forest-monitoring workshop.
01 — Presence
No invites, no accounts, no friction. A Driades room is just a URL. Send it to anyone — they open it and they're geo-anchored on the same map, zoom level and coordinates as the rest of the team.
You can literally point at something. Cursors carry names. Click an avatar and your camera glides along with theirs until you take the wheel back.
02 — Data
Drag PMTiles, Cloud-Optimized GeoTIFF, GeoParquet, GeoJSON or Zarr datacubes straight from your desktop. Or paste a URL to an S3 bucket — Driades reads data in place over HTTP range requests, so a gigabyte layer is on the map in seconds without touching a server.
Layers behave like Photoshop: reorder freely, tweak opacity per layer, lock the ones you're done with. URL layers are shared with the room automatically — everyone sees what you see the moment it loads.
03 — Clipping
Draw a boundary and anything you drop is clipped to that exact shape. Vectors filter to the polygon instantaneously. Rasters are genuinely masked — every tile composited in pixel space so the edges outside really do go transparent.
Tweak a vertex and every layer bound to that area re-clips, live, on every collaborator's screen. The boundary stays editable forever.
04 — Annotation
Click any shape for its geodesic area, perimeter and length — the numbers everyone in the room sees update the instant you draw. Name an area "Sector B" and it travels with the document.
Drop sticky notes pinned to exact coordinates. Two people can type in the same note at once. Every action lives in a shared, filterable history you can rewind with Ctrl+Z — or scrub back years and see what was said then.
05 — Time
Time is treated like any other dimension. One shared timeline everyone in the room moves together — drag the slider and observations carrying dates appear and disappear, drawings from that year vanish, notes reappear.
Stack satellite imagery by year and press play: a flipbook, a cross-fade blend, or a true per-pixel mean — all in sync for every collaborator watching.
06 — Analysis
DuckDB-WASM executes spatial SQL directly in your browser — query any loaded layer, filter by geometry or attribute, and mount the results straight back as a new coloured layer on the canvas. No server roundtrip.
For heavier work, pipe a drawn area into your own API or a hosted Gradio app. Inputs arrive prefilled, results land as layers, the full job history is shared with everyone in the room.
07 — Dryads
Scripts and AI agents join a room as full peers — we call them dryads, and each one moves a visible 🧚 cursor across the map. A Python client and an MCP server give them the same powers you have: draw, annotate, mount layers, set the time, fly the camera.
Everything a dryad does is recorded in the same document — you don't read a log, you watch it work, live, alongside your human team.
Made for real work
Driades grew out of workshops with ecologists, fire-recovery teams and field researchers who needed to share maps without learning GIS. Here is what they do with it.
Mount Sentinel-2 composites before and after the burn. Draw sectors, annotate regrowth density, run vegetation-index queries and export a PDF report the agency can read.
Drop LiDAR-derived canopy height tiles, plot ground-truth sample polygons, measure stand area and share the room with the sawmill contractor — everyone works off the same live dataset.
Stack yearly Landsat frames on the timeline, scrub backwards to compare crops, deforestation or glacier retreat. The sequencer plays it as a movie the whole team watches together.
Local activists map water sources, trails and hazard zones in a shared room. No accounts, no lock-in — just a public map anyone can open and build on.
A professor shares a room URL. Twenty students drop their own data, sketch hypotheses, query each other's layers and export their findings — all inside one browser tab.
Let a Python agent watch the canvas, classify land cover automatically, and draw polygons around detected features — while you review and refine each result in real time.
All of it ships in the open-source app, and all of it is free.
Points, lines, polygons, rectangles and freehand — geo-anchored and streamed live, including shapes still being drawn.
Stack as many layers as you like, with per-layer opacity, visibility and reordering.
Dropped files stream peer-to-peer to the whole room, survive reloads, and reach whoever joins later.
Geodesic area, perimeter and length on any shape, with shared names that travel in the document.
A shared log of every action, filterable by person, layer or area — click an entry to fly there.
See everyone's live cursor, and click an avatar to ride their camera across the map.
Play stacks of imagery as a flipbook, blended, or as a true per-pixel mean — in sync for the whole room.
Choose who you're time-synced with: co-edit a shared timeline, or follow someone read-only.
Point at an xarray-style store and scrub its time axis with the shared slider.
Elevation your data already carries extrudes into columns and elevated paths, with an exaggeration slider.
Deep-link layers by URL or push them live from a node editor over postMessage.
Workspace JSON, GeoJSON annotations, Markdown and PDF reports, PNG snapshots, video recordings.
Rotate the canvas without moving the data — it's a per-user view, so nobody else's map turns.
Your workspace persists locally and merges automatically when you reconnect.
Responsive layout with touch drawing, touch notes and finger-position presence.
No accounts, no tracking, no analytics baked into the canvas. Rooms can be end-to-end encrypted.
The commons
On the free tier, every shared map gets a public link and a place in the gallery. Browse, open, fork — build on someone else's work.
The canvas itself is free and open source. The paid tier adds the things that need someone else's servers: a curated data catalog, hosted AI, and maps that stay online when your browser is closed.
Free, forever
€0
The full collaborative canvas, running in your browser. Self-host it if you want to.
Your data stays on your machine and with your peers. Rooms persist in your own browser storage.
For teams that need it hosted
TBA/ month
Everything in Sapling, plus the infrastructure — data, AI and hosting we run for you.
Waitlist members get founding pricing and first access to the catalog.
| Sapling Free | Canopy Paid | |
|---|---|---|
| Collaborative canvas, unlimited rooms | ✓ | ✓ |
| All data formats & drop-to-clip | ✓ | ✓ |
| Spatial SQL, SPARQL, 3D, sequencer | ✓ | ✓ |
| Agent API (Python + MCP) | ✓ | ✓ |
| Self-hosting | ✓ | ✓ |
| AI assistant | Bring your own key, runs locally | Hosted chat, no key needed |
| Dataset catalog | Bring your own data | Curated catalog, one click |
| Map & gallery hosting | Needs a browser open | Always on, no browser required |
| Sharing | Public maps | Public and private maps |
| Room persistence | Your browser's local storage | Server-side snapshots & history |
| Support | Community & GitHub | Priority |
The canvas, the sync layer, the loaders, the agent bridge — all of it. You can clone it, run it on a laptop at a workshop with no internet, or deploy it inside your institution and never talk to us.
Sync happens peer-to-peer over WebRTC using Y.js CRDTs. The only shared infrastructure is a tiny stateless signaling server for peer discovery — it never sees your data, and rooms can be end-to-end encrypted on top. Data layers are read directly from S3-compatible storage.
The paid tier exists to fund the free one, not to fence it off. Nothing that works today will move behind the paywall.
All open-source licensed. No proprietary dependencies, no API keys to render a map.
From workshops
Real rooms used in forests, classrooms and crisis zones.
"We mapped three fire perimeters in an afternoon with a mixed team — foresters, data scientists and a fire chief who had never opened a GIS. Nobody needed to install anything."
"The timeline made our glacier retreat story visceral. Students scrubbed through 20 years of Landsat and drew polygons on the same frame — all in one browser tab."
"I connected our Python change-detection pipeline as an agent. It drew polygons, I reviewed them, and the whole session became a publishable dataset — without leaving the map."
"Community activists used it during the floods — a shared URL, drop a phone-photo layer, mark safe routes. No accounts, no barrier to entry. It just worked under pressure."
Who we are
Driades was born at h4ck1ng.science — a small research collective that builds open tools for people who work with land, data and each other. We run workshops in forests and classrooms, watch how teams actually collaborate, and ship software that removes the friction we see.
The canvas started as a single question: why does sharing a map still require a GIS degree and an IT department? Three years of prototypes, field tests and peer-to-peer experiments later, Driades is what we wish we had from day one.
No venture funding, no growth team, no exit strategy. Just a maintained open-source tool funded by the people who find it useful enough to pay for hosted extras.
Not public yet
Driades is being used in workshops today and opens to everyone soon. Join the waitlist and you'll be the first to know — plus founding pricing on Canopy and early access to the dataset catalog.
No. It's the actual open-source app, unlimited, with no expiry. The paid tier adds hosted services on top — it doesn't unlock the canvas.
You paste an API key from your own provider account into the app. It's stored in your browser and requests go directly from your machine to that provider — we never see the key or the conversation. On Canopy, we run the model for you instead, so there's nothing to set up and nothing to pay for separately.
Hosting a map that stays online costs something, and open geospatial work is worth more when others can see it. Public maps get hosted at no charge; private ones are the paid feature. You can always keep a map entirely to yourself by not sharing it — unshared rooms are peer-to-peer and never touch our servers.
Yes — npm run build produces a static bundle you can host anywhere, or bring the app, signaling server and agent bridge up together with docker compose up -d. The only moving part collaborators need is a reachable signaling server.
Not by default. Layers loaded by URL are read straight from wherever they live; files you drop are streamed peer-to-peer to the people in your room and cached in each browser. Nothing passes through a server unless you choose hosted maps.
Not decided yet — we want it to be affordable for research groups and small teams. Waitlist members get founding pricing and will hear the number before anyone else.
Explore
Tutorials, community canvases and curated datasets — all open to browse and build on.
Tutorials, walkthroughs and case studies from people using Driades in forests, classrooms and crisis zones.
Read the field notes →Live maps and dashboards built by the community — fire tracking, glacier monitoring, community atlases and more.
Browse the gallery →Curated geospatial data catalog — Sentinel-2, canopy height, elevation, land cover and more. One click to mount.
Explore the catalog →