Payload CMS Extended MCP. Query, inspect, and update content data via chat.

You're connecting your AI agent straight to your Payload CMS backend. This isn't just a wrapper; it lets you manage and manipulate all your content data—creating, reading, updating, or deleting records—without ever opening the actual Payload Admin panel. Your agent gets deep access to your entire data model through natural conversation.

Data Awareness: Knowing What You Got

You don't need to switch between your code editor and the CMS console just to verify a field name or check an ID number. The server lets you discover what data exists first. Use list_collections to see every collection defined in your schema, giving you a map of all available content types.

You can also call get_global to fetch system-level configuration and global settings stored outside of any specific collection.

To find documents, use list_documents. This tool lets you search across collections using plain language queries while applying complex filters, specifying sorting parameters, or managing pagination limits. If you need the complete picture for one document, call get_document with a unique ID; this action returns the full JSON payload, revealing every field and relationship associated with that specific entry.

Content Manipulation: Writing to Your CMS

The server gives your agent full write access. You can create brand new documents in any collection using create_document. When content changes, you modify existing records by calling update_document with the target ID and the new data payload. If a record is obsolete, you can permanently delete it using delete_document.

Beyond collections, you manage system-wide settings directly by running update_global, allowing your agent to change configuration values that affect the whole site.

Authentication and Context Control

Because these are sensitive operations, managing user context is key. To start working, you'll first use login_user with credentials; this action provides an active session token needed for all subsequent calls. You can verify who your agent is acting as by running get_me, which returns details about the currently authenticated user session.

When you're done or need to switch users, call logout_user to end the current session and invalidate tokens. If the access token expires during a long workflow, you won't be locked out; you just use refresh_token with your stored refresh token to grab a new, valid access token for continued operation.