Your agent works with your CMS.

Begin with an idea, a design or a change you want to see: “A bright hotel website,” “Add a new service,” or “Prepare an illustrated article.” The agent uses configured CMS tools to prepare something you can open and discuss.

The agent, model and CMS have different roles. The agent organizes tasks using a chosen model; Progressive CMS stores the website and provides editing, validation and publication operations. Agent access and external services are configured separately. Read Codex’s experience or the field guide for agents.

Codex, Claude Code and OpenCode can be configured for project work. Messaging and recurring workflows can be built around tools such as OpenClaw or Hermes Agent. These are separate products: available capabilities depend on their configuration and the CMS permissions you grant.

Your AI agent connects to Progressive CMS through a machine interface: REST API, CLI or MCP. The connection lets it carry out real CMS operations, including discovery, structured content work, validation and preview. The available actions depend on the installed interfaces and the access you give the account.

REST API connects HTTP clients and custom agent integrations to the CMS. CLI supports agents that work in a terminal and documented command-line workflows. MCP (Model Context Protocol) gives compatible assistants discoverable CMS tools with defined inputs.

Choose the interface your agent supports, configure the instance address and provide a dedicated service-account token through its private settings. Start with the CMS bootstrap and published contracts, then prepare a draft, validate it and review the actual page before publication.

You can ask for a new article, a change to a service page or a revised design. The agent performs the corresponding work through CMS operations and project tools. The result stays in Progressive CMS, where your team can edit it in the admin. Explore the CMS REST API, CLI and MCP interfaces.

The agent receives a service account and token with permitted operations. The owner can limit access, set an expiry and revoke the token. Content editing, theme work, preview and publication have separate permissions. People and agents use the same underlying CMS operations, so each team can work within its responsibility.

Prepare content in a draft and validate it. A private preview shows the page with real data; a candidate theme lets you assess a design before activation.

Publication is a separate action for an authorized participant. Changes become part of the CMS history, rather than living only in a conversation.

Editors can update content in the admin while an agent prepares new articles, sections and changes. A studio can develop the theme independently. The published site works without an AI model attached to every visitor request; the agent connects when there is work to do.