Getting started

What is Showly

Agent-native hosting for websites: connect a repo, point an agent at it, ship through previews.

Showly is a website host built around the assumption that the person writing the change is an agent — Claude Code, Codex, or any MCP-aware coding agent — and the person approving the change is a human on your team. The platform is designed around that split.

The shape of a Showly workspace

A Showly workspace contains one or more sites. A site is a Git repository (today, GitHub) plus the build, deploy, and runtime configuration Showly derives from it. Every change to a site flows through the same path:

  1. An agent reads the site's context via MCP tools.
  2. The agent produces a _change plan_ — a description of what files it will modify and why.
  3. The plan is applied as a patch in an isolated workspace; Showly builds and produces a preview URL.
  4. A human approver reviews the preview, the diff, and any check results.
  5. Approval unlocks a production publish; a failed approval simply leaves the preview in place.

No matter the channel — chat, MCP tool call, or REST API — every change rides this same rail.

Where Showly is opinionated

  • Previews are mandatory. Production never receives a change that hasn't lived as a preview first.
  • Agents never hold production keys. Tools are scoped per-site, per-action, and audit every call.
  • Rollback is one click. Every publish leaves an immutable artifact you can revert to.
  • Audit is non-negotiable. Every action — agent or human — produces an audit record.

Where Showly is flexible

  • You bring your own Git host (GitHub today; GitLab and Bitbucket on the roadmap).
  • You bring your own agent. Showly publishes a Skill for Claude Code and an MCP server any agent can target.
  • You pick your runtime. Static CDN and Cloud Run (serverless containers) are the managed targets today, with bring-your-own-cloud deploy targets on supported plans.

Next steps