Prezly MCP

Prezly MCP lets an AI assistant find information and carry out supported tasks in your Prezly account. You can ask about stories, contacts, campaigns, pitches, coverage, and newsroom settings without building an API integration yourself.

Your assistant needs a client that supports a remote Streamable HTTP server with bearer-token authentication.

Connect your assistantCopy Anchor Link

1. Create a personal API token and configure your client. Start with read permissions for the data you need.
2. Enable Prezly in the conversation and try a read-only task.
3. Review the result before giving the assistant permission to make changes.

The hosted server is:

https://mcp.prezly.com/mcp

A listed tool does not prove that your token can use it. Access depends on the token's scopes, your current Prezly permissions, newsroom access, and whether the underlying endpoint accepts personal tokens. Adding every scope does not unlock every tool.

Try your first read-only requestCopy Anchor Link

After connecting, ask:

List my Prezly newsrooms. Show their names and identifiers. Do not change anything.

This tests a real read, rather than just tool discovery. Include sites:read on the token. You should see the newsrooms available to the connected user and account.

Then try a specific request, with stories:read as well:

In the newsroom called Acme, find the five most recently updated draft stories.
Show each exact title, identifier, and update time. Do not retrieve full content or make changes.

Replace Acme with a returned newsroom name. If there are similarly named newsrooms, use its returned identifier. An empty result can be valid; it does not necessarily mean the connection failed.

Give the assistant a bounded taskCopy Anchor Link

State which records to use and what outcome you want. For example:

Find the draft story called Acme introduces its new collection in the Acme newsroom.
Read it and suggest a shorter subtitle. Show me the suggestion before changing the draft.
Do not publish, schedule, send, or delete anything.

A draft update needs the relevant read and write scopes. The assistant should identify the story from search, read its current state, and use that state when submitting an approved change. Publishing and sending require separate permissions; creating or editing a draft does not implicitly publish or send it.

Keep your token in the client's private credential settings. Tool results can contain unpublished content and personal data, so choose an assistant and workspace appropriate for that information.

Continue with MCPCopy Anchor Link

• Set up MCP: create a token, connect a client, and verify access.
• MCP tools: choose a workflow and check personal-token restrictions.
• Safety and privacy: control changes and understand data handling.
• Troubleshooting: diagnose failures and reconcile uncertain outcomes.

For REST integrations and custom websites, start with authentication and your first API request.