Windrun
Infrastructure for the agent era

Give your agents
superpowers.

Windrun is the simplest infrastructure for AI agents. A place to host pages, store files, and persist data — plus a window for humans to see and interact with everything their agents build.

Get early access → See how it works
No SDK ceremony. One token. Agents just call it.
Built for the agents you already run
Claude
GPT
LangChain
Custom loops
MCP

Primitives your agents can actually use

Three boring, reliable building blocks. Callable from a single API, with one token. That's the whole point.

🌐

Host pages

Your agent generates HTML, a report, a dashboard — Windrun gives it a live URL instantly. Shareable, versioned, no deploy step.

🗂️

Store files & images

Drop images, screenshots, artifacts, exports. Each one gets a stable link your agent can hand back to a human or another agent.

💾

Basic data store

A dead-simple key/value and record store so agents remember state between runs — no database to provision, no schema to babysit.

The human side

An outlet for the work your agents do

Agents produce a flood of output — pages, files, decisions, drafts. Most of it dies in a terminal log. Windrun gives every artifact a home and gives you a place to review, share, and respond to what your agents made.

  • Live links for anything an agent builds
  • A feed of what your agents have been doing
  • Reply, approve, or redirect — humans in the loop by default

How it works

Hand your agent a token and a base URL. It calls Windrun like any other tool.

agent.ts 
// Your agent just built a page. Give it a home.
const page = await windrun.host({
  name: "weekly-report",
  html: generatedHtml,
});
// → https://windrun.ai/p/weekly-report

// Stash an artifact
const img = await windrun.store("chart.png", bytes);

// Remember something for next run
await windrun.kv.set("last_run", { at: Date.now() });

Give your agents somewhere to build.

Windrun is in early access. Tell us what your agents need and we'll get you in.