Connect your agent

Pick your client

Every client below points at the same remote MCP server. There is no key to paste and no sign-in step: deploy tools work anonymously, and you only log in later if you want to claim a site and keep it.

Claude Code (plugin)

The plugin bundles the deploy skill and the MCP connection, and updates itself — you never reinstall it to get new guidance.

/plugin marketplace add aft-page/plugins
/plugin install aft-page@aft-page

Restart your Claude session so the MCP server and tools take effect, then ask “Deploy this to aft.page”. In auto mode Claude Code won't let an agent install a plugin on its own, so type it yourself. Use the plugin or the claude mcp add line below, not both — the plugin already includes the server.

Any agent with the plugins installer

npx plugins add aft-page/plugins

One install for Claude Code, Cursor, Codex, Copilot CLI, VS Code, Grok, and Kimi. Restart the agent, then ask it to deploy.

Claude Code · Cursor · Windsurf (MCP by hand)

Claude Code writes the config for you:

claude mcp add --transport http --scope user aft-page https://mcp.aft.page/mcp

By hand the entry is the same everywhere — Claude Code ~/.claude.json (user) or .mcp.json (project), Cursor ~/.cursor/mcp.json, Windsurf ~/.codeium/windsurf/mcp_config.json:

{
  "mcpServers": {
    "aft-page": {
      "type": "http",
      "url": "https://mcp.aft.page/mcp"
    }
  }
}

Windsurf takes serverUrl instead of url. There is no headers block — no token, no OAuth app to create. Reload the host, then ask “Deploy this to aft.page”.

Claude.ai · ChatGPT · Grok — connectors

Paste the URL and nothing else:

https://mcp.aft.page/mcp

• Claude — Settings → Connectors → Add custom connector.
• ChatGPT — turn on Developer mode (Settings → Security and login), then add the URL as a custom server.
• Grok — grok.com/connectors → New connector → Custom.

Start a new conversation after adding it so the tools load. Deploys from a connector land on a live URL like any other; keep one by opening its claim link and attaching it to an account.

No MCP — CLI

curl -fsSL https://aft.page/install | sh
aft deploy

No login to ship. Agents adding --json get one result object, no prompts, exit 0 = live. CLI guide.

Deeper per-client notes live under /with/claude, /with/cursor, /with/codex, and /mcp.

Guides

Once your agent is connected, the same deploy paths work by hand. Same result: https://{slug}.aft.page.

API (one-shot)

curl -X POST https://api.aft.page/v1/deploy \
  -H 'Content-Type: text/html' \
  -H 'X-Aft-Client: curl' \
  --data '<h1>Hello from aft.page</h1>'

Multi-file: JSON { "files": [{ "path", "content" }] } or multipart. Optional ?slug=vite-hello. A collision gets a numeric suffix — an existing slug is never overwritten by a new, unrelated deploy. Full schemas: /mcp.md.

What we support

aft.page does not run npm run build. You (or your agent) detect the project, build locally if needed, then upload ready files.

Never upload src/, node_modules, or .next/. Always include index.html for static sites. Per-framework build steps, verified vs. untested frameworks, and what still needs a runtime we don't run yet: Frameworks guide.

Logs

Owner and editors: project page → Logs, or GET /v1/sites/{slug}/logs. Request-level hits and errors for the live site, plus recent deploy failures. No IP addresses. Kept 7 days.

{
  "slug": "my-app",
  "views": { "today": 12, "d7": 340 },
  "lastServedAt": "2026-08-12T09:14:02.000Z",
  "events": [
    { "createdAt": "2026-08-12T09:14:02.000Z", "method": "GET",
      "path": "/api/widgets", "status": 500, "bytes": 214, "country": "US" }
  ],
  "deployFailures": [],
  "observability": {
    "window": "12h",
    "totals": { "ok": 40, "fourxx": 2, "fivexx": 1, "bytes": 12000, "requests": 43 },
    "buckets": [],
    "countries": [{ "country": "US", "n": 30 }, { "country": "IN", "n": 13 }]
  }
}

events is capped at the most recent 100 rows; only document-shaped hits and non-2xx responses stick around — noisy favicon/robots-style paths are filtered before they're stored.

Source

Owner, editors, and viewers: project page → Source, or GET /v1/sites/{slug}/files. File list for the current deploy (or an older one via ?deployId=); click a text file to preview its contents. This is a private read of what you uploaded — it's not a public URL, and the live site is unchanged by looking at it.

{
  "slug": "my-app",
  "deployId": "dep_abc123def456",
  "files": [
    { "path": "index.html", "bytes": 1820 },
    { "path": "assets/app.js", "bytes": 44210 }
  ]
}

Limits

Live for 60 minutes. Claim within 7 days to keep it. Claim guide.

Examples

• hello.aft.page — plain HTML
• vite-hello.aft.page — React + Vite dist/
• next-hello.aft.page — Next via upstream
• share-checklist.aft.page — static multi-file