Terabytes of footage, projects and images that open instantly and take no disk space. Every save is a version, AI transcribes and tags files as they land, and your agents work on the same drive.
Native macOS app · Web · Linux mount and CLI · Windows coming later
Open and edit everything in a space right from Finder. Files you haven't opened are placeholders, so they never fill up your Mac.
A001_C003.mov
Modified: Today, 4:12 PM
Add Tags…
General:
More Info:
Name & Extension:
Comments:
Side Space streams files to the app that asks for them. Editing, design and 3D apps just see ordinary files, so nobody ships drives or waits hours for downloads.
Press ⌥ Space to search names, AI tags, transcripts and prompts across every space, and jump to the second a line was said. Let agents work with files that aren't on your disk. On the Mac, Side Space files show up in Spotlight too.
Always a keystroke away. Hits come from names, tags, transcripts and comments, and drag straight onto a timeline.
Turn on automations for a folder and every new file is transcribed, tagged and given a proxy. Results live in sidecars next to each version, never inside the file, and each one records who made it with which model.
{ "transcript": { "language": "en", "segments": [ { "start": 134.2, "end": 136.9, "text": "We shot the dock right at dusk." } ], "model": "whisper-1" }, "ai.tags": { "tags": ["interview", "dock", "dusk", "wide"], "model": "claude-opus-5-5" }, "media": { "duration_s": 642.5, "video_codec": "prores", "width": 3840, "height": 2160, "fps": 25 }}Any OpenAI-compatible speech API, in the cloud or on your own model server.
Claude or a local vision model describes what's on screen as searchable tags.
ffmpeg presets make deliverables and proxies, saved back to the space.
The same content with the same settings is processed once, and the result is reused.
One drive for your team, your tools and your agents.
Edit from anywhere and stop shipping drives. Review, versions and transcripts sit next to the footage.
Prompts, seeds and models travel with every version, so you always know how an image was made.
"ai.generation": {
"prompt": "portrait,
warm rim light",
"seed": 4127730918,
"model": "flux-1.1"
}The brand kit is always the latest, older versions are a click away, and partners get files by link.
v14
v13
v12
Give agents a persistent drive. Their changes show up for the team live, or try them first in a zero-copy fork.
side.fork_space ("Harbor Film")
→ 1.3 TB · 0 bytes copied
side.write_text ("notes.md")
Run it on your own server, or put spaces in the S3-compatible buckets you already have. Access goes down to teams and folders, and every sign-in and change is on record.
Numbers our acceptance tests hold us to
Agents connect over MCP, pipelines react to every change, render nodes mount the spaces, and your scripts use the same API as the apps. Each one gets a token that grants only what it needs.
Claude and other agents search, read, edit, organize and share files, and run AI actions, through /mcp.
Copy a multi-terabyte space instantly for an agent to try things, then apply what works.
Signed, ordered and retried for every change, or run an AI action on every upload that matches.
Linux render nodes and servers mount spaces as a folder; aws cli, rclone and boto3 read and write them too.
TypeScript, Python, Go and Rust, checked against the same 34 steps, plus OpenAPI 3.1 and llms.txt.
Limited by permission, space and expiry. Shown once; the server keeps only a hash.
import { SideSpace } from "@sidespace/sdk"; const side = new SideSpace({ baseUrl: "https://side.example.com", token: process.env.SIDESPACE_TOKEN }); // Search names, AI tags, transcripts and prompts.const hits = await side.search("harbor night", { type: "video" }); // Transcribe once; the same bytes are never billed twice.const run = await side.run("Film A", "/Ep01/a.mov", "side.transcribe"); // Tag every new clip in /dailies as it lands.await side.createPipeline("Film A", { name: "tag dailies", action: "side.tags", path_glob: "/dailies/**", mime_prefix: "video/",});Every plan is the whole product. Plans differ only in how much you store and how much AI you run.
For trying Side Space
Free
For one person on all their computers
$15/ month
$180 billed yearly · save 25%For teams working in shared spaces
$30/ member / month
$360 per member billed yearly · save 40%For companies with their own storage and compliance needs
Every plan includes
Still have a question? Write it in the early access form. A person reads every one. Get early access
A shared cloud drive for teams. On a Mac it's a location in Finder: files open on demand, every save is a version, and AI and agents work on the same drive. There's a web app for any browser too.
Both. Individuals use it to work on the same files from several computers; teams share spaces with member roles, organizations and team permissions.
Cloud drives sync whole files before you can use them. Side Space fetches only the parts an app reads, and you can open a file a teammate is still uploading. Every save is a version, AI metadata travels with versions, and agents have their own identity and permissions.
Files appear in Finder as placeholders, and opening one fetches only the chunks that are read. What you've read goes into a local cache with a size limit that drops the oldest first. Folders you want to take along can be made available offline, and you see the size first.
Yes. To Mac apps it's an ordinary folder, so editing, design and 3D apps open and save as usual. Apps that save by writing a temporary file and renaming it still make a new version of the same file, keeping its history and comments.
The last save becomes the current version and the other person is told right away. Their save is kept as a version they can restore in one click.
Files you've opened and folders you've made available offline work without a connection. Changes made offline sync when you're back.
Opening a file fetches only what's needed, so an ordinary connection gets you watching. For smooth scrubbing of high-bitrate media like 4K ProRes, faster is better, or make the folder available offline ahead of time.
Only when someone in the space runs an AI action or turns on an automation, and only the model service your server is set up with. That can be a cloud API or your own local model.
Yes. An agent joins a space as a member, and its token limits what it can read and write. It works through remote MCP, the HTTP API or the S3 API, and can try changes in a zero-copy fork that you review before applying.
Yes. TypeScript, Python, Go and Rust SDKs that pass one shared test suite, plus OpenAPI 3.1, a CLI, webhooks and an llms.txt for AI. For now they install from the repository; registry packages come later.
A native app on the Mac, with a location in Finder and files in Spotlight; the web app in any browser; a Linux mount for render nodes and servers; and a CLI for macOS and Linux. Windows is planned.
Files are split into chunks and encrypted with AES-256-GCM under a key per space before they're stored, so the bucket holds only ciphertext. Members are viewers, commenters, editors or managers, and share links can have passwords and expiry dates and be turned off at any time.
Yes. Any space can live in your own S3-compatible bucket, such as AWS S3, Cloudflare R2, Backblaze B2, Alibaba Cloud OSS or your own MinIO. You can also run the whole service on your own server.
We're opening up in batches. Tell us what you're working on and we'll email you when it's your turn.