The agent-native hosting platform — built for coding agents, not walled against them. Deploy any folder to a live URL in one command — no account, no dashboard, no CAPTCHA:

npx openpouch deploy

Status: technical preview. openpouch deploy serves static sites (HTML/SPAs/build output) and runs real Node.js apps in a hardened container — autonomously, from any agent or terminal. Previews are ephemeral (unclaimed ones vanish after 72 h); durable production hosting and self-service billing are on the way. Feedback welcome on GitHub.

You get a live https://<slug>.openpouch.sh URL plus a claim link. The agent deploys autonomously; a human claims it via the link to keep it. openpouch writes the deployment truth (deploy.manifest.json, deploy.evidence.json, DEPLOYMENT.md) back into your repo, so any agent can resume after context loss.

Framework frontends (React/Vite/Next/Svelte…): deploy the built output, not the source folder. Build first, then point deploy at the build directory:
npm run build
npx openpouch deploy dist     # Vite → dist/ · CRA → build/ · Next static export → out/
If you run openpouch deploy in an unbuilt frontend root, it stops and tells you which folder to deploy instead of silently shipping source that won't run (use openpouch deploy . to force the current folder). Server-side build-on-deploy is live — pass the source folder explicitly (openpouch deploy .) and the server runs npm run build for you, for both full-stack apps (an Express/Node server with a build step) and unbuilt static SPAs (a Vite/CRA frontend with no server → built, then served as files).

Accounts (optional)

Start anonymous, or create a free account for higher limits — entirely from the agent, no dashboard:

npx openpouch signup --email you@example.com    # or: --github
npx openpouch activate --account <id> --token <token-from-email>
npx openpouch whoami --json                      # your tier + current usage

Your API key is stored locally; deploys then run under your account and quota. Abuse is controlled with accounts, quotas, rate limits and an egress filter — never by asking a machine to prove it's human.

Bring your own provider

Already on Render or Vercel? openpouch init detects your project and maps the service; then openpouch preview / openpouch prod run a governed pipeline — previews autonomous, production gated behind a human approval. Agents can never self-approve production.

Commands

deploy · list · delete · signup · activate · whoami · init · inspect · plan · preview · prod · approve · verify · logs · rollback · feedback — every command supports --json and stable exit codes, so any agent harness with a shell can drive it. An MCP server exposes the same tools.