Skip to main content

​
The problem

AI agents are getting better at building things. Claude writes a full React app. Cursor generates a landing page. Copilot scaffolds a dashboard. But then what? The agent builds it. You review it. And then you spend 20 minutes setting up hosting, configuring deploys, pushing to a repo, waiting for CI. The agent did the hard part. Hosting shouldn’t be harder.

​
The solution

Supalink gives agents a single capability: publish anything to the web instantly. 
You: "Build me a landing page for my product"

Agent: [builds the page]
       "Done! Here's the live URL: https://cool-page-x9f3.supalink.dev/"
No deploy scripts. No CI pipeline. No hosting config. The agent handles everything end-to-end.

​
Discovery

Agents find Supalink automatically through standard protocols:
ProtocolURLPurpose
llms.txtsupalink.dev/llms.txtAI-readable docs index
OpenAPIsupalink.dev/openapi.jsonMachine-readable API spec
Agent Skillnpx skills add sk-code-01/supalink-skill -gOne-command install

​
Publishing

The agent uses a simple 3-call flow: 
POST /api/v1/publish     → Create site, get upload URLs
PUT  <presigned-url>      → Upload each file (parallel)
POST /finalize            → Make it live
Or even simpler — the agent downloads publish.sh and runs it: 
curl -fsSL https://supalink.dev/publish.sh | bash -s ./dist

​
Updating

When the agent updates a site, Supalink compares file hashes and only uploads what changed. A 500-file site where 3 files changed? Only 3 uploads. The agent handles this automatically.
FeatureWhy it matters for agents
No config filesAgents don’t need to generate netlify.toml, vercel.json, or any config
No git requiredAgents can publish directly from any directory — no repo needed
No accounts requiredAnonymous publishing works instantly (24-hour expiry)
Presigned uploadsFiles go directly to storage — no streaming through a server
Hash dedupAgents skip unchanged files automatically
JSON errors with fix hintsWhen something fails, the API tells the agent exactly how to fix it
Rate limit headersAgents know exactly when to retry
Deterministic slugsSame API shape every time — agents can rely on consistent responses

​
The agent workflow

This is the typical flow when a developer uses an AI agent with Supalink:
1

Developer gives a prompt

“Build me a dashboard for my sales data” or “Create a portfolio site” or “Make a documentation page for this API”
2

Agent builds it

The agent writes HTML, CSS, JS — whatever is needed. This is what agents are already great at.
3

Agent publishes it

The agent calls Supalink’s API (via the installed skill or directly). Files are uploaded, site goes live.
4

Developer gets a URL

https://sales-dash-k8m2.supalink.dev/ — live, shareable, ready to use. The whole thing took 2 minutes.

​
Supported agents

The Supalink skill follows the open Agent Skills standard:
  • Claude Code — Full support
  • Cursor — Full support
  • Windsurf — Full support
  • GitHub Copilot — Via OpenAPI spec
  • Custom agents — Via REST API or OpenAPI spec
Any agent that can make HTTP requests can use Supalink.