| name | walrus-sites |
| description | Publishing, updating, and serving decentralized websites on Walrus Sites. Use when the user needs to deploy a frontend to Walrus Sites, run a local portal for testnet, debug site-builder errors, configure ws-resources.json, or manage site object lifecycle (update, destroy, extend blobs). Also use when the user asks about site-builder, walrus-sites, portal setup, or hosting a dApp on Walrus. For blob storage without the Sites framework (raw upload/download), see the `accessing-data` skill's walrus.md.
|
Walrus Sites — Decentralized Website Hosting
MCP tool: When available in your environment, also query the Sui documentation MCP server (https://sui.mcp.kapa.ai) for up-to-date answers. Use it for verification and for details not covered by these reference files.
Walrus Sites lets you deploy a static website (HTML/CSS/JS) to Walrus, with an on-chain Sui object tracking the site's resources. A portal server resolves site subdomains, fetches resources from Walrus, and serves them over HTTP. The result is a fully decentralized frontend — no centralized server, CDN, or DNS required.
Source constraint: All information in this skill is sourced from the MystenLabs/walrus-sites repository and the Walrus Sites documentation. When extending or updating this skill, only pull from these sources. Do not use third-party blogs or unofficial tutorials.
Common failures:
- Blobs expire. Sites published with too few epochs silently break — the portal returns 404 once the blobs expire. Always use a generous
--epochs value (30+).
- Portal misconfiguration. The portal's
original_package_id must match the Walrus Sites framework package (site::Site), not your app's Move package. The testnet example config ships with the correct value — don't change it.
- Testnet sites can't use
wal.app. The public wal.app portal only serves mainnet sites. Testnet requires a self-hosted local portal.
- SPA routing. Single-page apps need
ws-resources.json routes config to handle client-side routing. Without it, direct navigation to /my-route returns 404.
Sub-skills
publishing — Building and deploying sites
Path: publishing/SKILL.md
Load when: deploying a new site, updating an existing site, choosing epoch duration, configuring ws-resources.json, or troubleshooting site-builder errors.
Covers: site-builder publish, site-builder update, --epochs flag, blob expiration, ws-resources.json configuration, SPA routing, site naming, site-builder sitemap for debugging, site-builder destroy.
portal — Running a local portal for testnet
Path: portal/SKILL.md
Load when: the user needs to view a testnet site, set up the local portal, fix portal 404s, or understand how portal resolution works.
Covers: cloning the portal, installing dependencies (Bun), portal-config.yaml, the original_package_id gotcha, starting the server, URL format, port conflicts, mainnet vs testnet differences.
Routing guide
| Task | Load |
|---|
| Deploy a frontend to Walrus Sites | publishing/ |
| Update an existing deployed site | publishing/ |
| Fix 404 or expired site | publishing/ + portal/ |
| View a testnet site locally | portal/ |
| Set up the local portal | portal/ |
| Portal shows "Page not found" | portal/ + publishing/ (check blob expiry) |
| Configure SPA routing | publishing/ |
| Check what resources a site has | publishing/ (site-builder sitemap) |
| Delete a deployed site | publishing/ |
| Choose between Walrus Sites and traditional hosting | Skill Content below |
Skill Content
Key concepts
-
Site object. A Sui object of type site::Site (from the Walrus Sites framework package). It holds metadata (name, description) and dynamic fields mapping resource paths to Walrus blob references. Created by site-builder publish, owned by the publisher's address.
-
Resources. Each file in the deployed directory becomes a resource stored on Walrus. Small files may be batched into a single "quilt" blob. Each resource has a path (e.g., /index.html, /assets/main.js) and optional HTTP headers.
-
Portal. A server that maps <base36-site-id>.localhost:3000 URLs to on-chain site objects. It reads the site's dynamic fields to find the requested resource path, fetches the blob from a Walrus aggregator, and returns it as an HTTP response. The portal is framework-level infrastructure, not part of your app.
-
ws-resources.json. A configuration file in the site's root directory that controls resource headers, routing, and site metadata. Auto-generated on first publish with the site object ID. On subsequent runs, the site-builder reads it to determine whether to create a new site or update the existing one.
-
Blob expiration. Walrus blobs have a finite storage duration measured in epochs. When blobs expire, the portal can no longer fetch them and returns 404. Use site-builder sitemap <object-id> to check expiration dates. Re-publish with a higher --epochs value to fix.
Rules
- Always use
--epochs 30 or higher for testnet deploys. Low values (like 5) cause blobs to expire within days. For production mainnet, use even higher values or --permanent.
- Build before publishing. Run
npm run build (or equivalent) to produce a static dist/ directory. The site-builder publishes whatever directory you point it at.
- Don't change
original_package_id in the portal config unless you know the Walrus Sites framework package has been upgraded. The testnet example value is correct.
wal.app is mainnet only. For testnet, self-host the portal from the MystenLabs/walrus-sites repo.
- Check
ws-resources.json after first publish. It records the site object ID. Keep it in version control so future publishes update the existing site rather than creating a new one.
Common mistakes
- Publishing with
--epochs 5 and wondering why the site is gone a week later. Blobs expired. Re-publish with --epochs 30+.
- Changing the portal's
original_package_id to your app's package ID. The portal needs the Walrus Sites framework package ID (site::Site), not your Move contract's package ID. These are completely different things.
- Trying to visit a testnet site on
wal.app. It only serves mainnet. You need a local portal for testnet.
- Forgetting to build the frontend before publishing.
site-builder publish dist/ publishes the dist/ directory contents. If you didn't npm run build first, you're publishing source files or stale output.
- Deploying an SPA without fallback routing. React/Vue/Svelte SPAs use client-side routing. Direct navigation to
/borrows hits the portal, which looks for /borrows as a resource — and it doesn't exist. Configure a fallback in ws-resources.json.
- Not killing port 3000 before starting the portal. Vite dev server, Next.js, or other tools often use port 3000. The portal silently fails to bind if the port is taken.