Product-led growth depends on one thing: how fast someone can go from nothing to a working result.
For over a decade at Cloudinary, that someone was a developer with a free account and a few lines of code. Today, it’s just as often an agent working on that developer’s behalf.
But an agent explores differently. It doesn’t read a landing page or compare feature tables. It wants to try it: Spin something up, build against it, run the result, and decide for itself, right now, without waiting on anyone.
So we removed the one step that couldn’t happen without a human: creating the account. No signup, no email confirmation, no click to wait on mid-build.
One command gives any agent a complete, ready-to-use Cloudinary environment, live from the first call, without requiring an email, payment, or ToS click.
npx @cloudinary/cloud
→ cloud_name: … claim_url: … expires_at: 24hCode language: CSS (css)Upload, transform, deliver, and preview your results the moment the command returns. The full pipeline is live from line one:
- Content-aware cropping (
g_auto). - Automatic format and quality (
f_auto, q_auto). - Background removal and generative fill.
- Video transcoding.
- Generating images from other images or a prompt, and converting stills into video.
The media plumbing — provisioning, building and verifying work — is the agent’s job now. Yours is to ship the app.
Cloudinary wasn’t retrofitted for agents. It’s been API-first since day one, built from the ground up for developers who’d rather integrate than click. That same foundation is what makes it native to agents. We have an MCP server agents call directly, and they already are. Agents are provisioning, uploading, and transforming through it today.
There’s also a reason an agent reaches for us instead of rolling its own media pipeline. It can write a resize in a few lines of sharp, but it can’t stand up a global delivery network, real-time transformation at scale, AI transforms that run in the delivery path, or durable storage and media governance on its own.
The Claimable Cloud gives the agent the whole platform immediately, at full production capacity. Claiming it unlocks durability and delivery reach: storage that doesn’t expire, global delivery, and governance over what’s public.
You don’t have to take our word for it. Run one line:
npx @cloudinary/cloudCode language: CSS (css)The response will include your cloud name, API key and secret, and a clam URL. For example:
CLOUD_NAME=a1b2c3d4
API_KEY=123456789012345
API_SECRET=«redacted»
CLOUDINARY_URL=cloudinary://123456789012345:«redacted»@a1b2c3d4
EXPIRES_AT=2026-08-06T08:29:07Z
CLAIM_URL=https://res.cloudinary.com/users/agent_email_confirmation?token=«redacted»
DELIVERY_IPS=203.0.113.7,198.51.100.42
ENV_FILE=./.envCode language: JavaScript (javascript)DELIVERY_IPS is the allowlist. Your assets deliver only to these IP addresses. If you encounter any issues with ACL restriction, claiming the cloud lifts the restriction.
Here’s a quick demo:
Then point your agent at something real. Here are some examples:
- A new page, built media-first. Build a performant landing page — hero video, image gallery, product shots — with content-aware crops that keep the subject centered from hero to thumbnail, format and quality optimized per browser, backgrounds removed and replaced, a video transcoded with its poster frame taken from the same source.
- Migrating an application you already have. Point your agent at an existing site and ask it to migrate the imagery: Find all referenced assets, upload them, and rewrite the references to optimized delivery URLs with responsive widths, automatic format and quality, and smart cropping.
- One source asset, multiple channels. Give your agent a campaign photo or video and the list of output channels for distribution. Produce the set: each aspect ratio cropped, each format and file size matched to its destination, a short vertical cut of the video. What comes back is the full variant set, and a pipeline that can also regenerate it when the source changes.
The Claimable Cloud is capped and time-boxed by design, but not simulated. Underneath it is the same infrastructure that serves our production traffic. The limits are on duration and reach, not on what the platform does.
While it lives, the agent has the full API and can preview its work in allowlisted environments — everything the build loop needs, and nothing to rebuild when a human claims it.
What the agent doesn’t get is a durable public URL. Publishing is a human’s call, on purpose. Agents get the freedom to explore, but humans keep accountability for what reaches the world.
When you like what the agent built and want to ship it, claim the cloud with your email. Everything your app needs transfers intact:
- Your images and videos.
- Variations and transformations.
- Your integration setup.
Same environment, now yours.
Point your Claimable Cloud storage to the bucket you already use and take it from there. You still get every media operation and the AI transforms on top. Lazy migration, no starting over.
Read more on this in our docs, point your agent to llms.txt, or run npx @cloudinary/cloud yourself.
The builder may have changed, but the principle hasn’t: Get to value fast, without anything in the way. Let the agent build something real. Then you decide what’s worth keeping.
