Agent-native deploy

Your agent writes it. It ships it live.

Install the EdgeSpark skill once. Then just tell Claude Code, Codex, Gemini CLI, or Cursor to deploy — it provisions a database, auth, and storage and returns a live URL. No dashboard, no config files.

Install the skill onceTell your agent to deployLive URL in seconds
See plans

What the skill gives your agent

Deploy without leaving the chat

Your agent runs the deploy itself through the EdgeSpark skill. You never open a dashboard or copy a wrangler command.

A real database, provisioned for you

Schemas are declared in TypeScript; the agent runs migrations. Your app saves real data from the first deploy.

Login that just works

Managed OAuth and email auth via configuration — your agent wires it up, no auth code to write or secrets to juggle.

Works with the agent you already use

First-class skills and plugins for Claude Code, Codex, Gemini CLI, Cursor, and Copilot — plus a generic skill + MCP path for anything else.

Served from the edge

Apps run on a global edge network, close to your users, with no servers for you to manage.

From idea to live in seconds

The whole loop — build in your agent, deploy, get a URL you can share — happens without a context switch.

How to use it

Three steps to a live app.

  1. 1

    Install the skill

    Paste one line into your agent — it installs the CLI and the skill on its own.

    ›Fetch and follow the guide on https://cdn.edgespark.dev/onboarding/install.txt
  2. 2

    Sign in & pay

    Unlock the live URL your agent deploys into — less than a coffee.

    $1/mo · Starter plan

  3. 3

    Tell your agent to deploy

    Say it in plain words. You get back a live URL — with a database, login, and storage.

    ›deploy this

Launch plan

Pick a plan, then ship.

Choose a paid workspace and unlock the live URL your agent deploys into. A free tier is available later from the dashboard.

Starter

$1/mo
Projects per organizationUp to 5
Database?Relational data per project — user accounts, posts, orders, etc. Good for most early-stage projects.Unlimited during Beta100 MB
Storage?Files, images, and videos per project. Upgrade when serving lots of product images, media uploads, or asset-heavy content.Unlimited during Beta512 MB
SupportCommunity
Get started

Pro

Most popular
$5/mo
Projects per organizationUp to 50
Database?Relational data per project — user accounts, posts, orders, etc. Good for most early-stage projects.Unlimited during Beta512 MB
Storage?Files, images, and videos per project. Upgrade when serving lots of product images, media uploads, or asset-heavy content.Unlimited during Beta256 GB
SupportPriority
Get started

Why agent-native matters

Most deploy platforms assume a human in a dashboard: you click through projects, paste config, wire up a database, set environment variables. That model breaks the moment an AI agent is doing the building, because the agent has to stop and hand the boring infrastructure work back to you.

EdgeSpark closes that gap. The deploy itself is exposed to the agent as a skill, so the agent that wrote your code is the same one that ships it. You stay in the conversation; the live URL comes back as a result.

That is the difference between "my agent built something on my laptop" and "my agent shipped something my users can open."

Frequently asked questions

How do I install the EdgeSpark skill?
Paste one line into your agent: “Fetch and follow the guide on https://cdn.edgespark.dev/onboarding/install.txt”. It installs the EdgeSpark CLI, signs you in, and adds the skill for your agent. Claude Code, Codex, Gemini CLI, Cursor, and Copilot are all detected automatically.
Do I need to know how to deploy?
No. That is the point of an agent-native platform — you tell your agent what you want and it handles provisioning, migrations, and the deploy. You get back a live link.
What does my agent deploy, exactly?
A full-stack web app: a frontend plus a backend with a managed database, authentication, and object storage. It runs on a global edge network.
Which agents are supported?
Claude Code, OpenAI Codex, Gemini CLI, Cursor, and GitHub Copilot have first-class integrations. Any other MCP-compatible agent can use the generic skill + MCP path.