MCP server

The Cevoid MCP server connects an AI assistant, such as Claude, ChatGPT, or Cursor, to your Cevoid workspaces. MCP (Model Context Protocol) is an open standard that AI assistants use to connect to other services. Once connected, you can ask the assistant about your analytics in plain language, have it create and update reports and dashboards, and have it edit your BrandKit.

Navigate to Settings → AI & Agents → MCP⁠ to manage the AI assistants connected to your workspace.

The MCP server gives an AI assistant access to Cevoid analytics and your BrandKit. You connect the assistant with your own Cevoid account, and choose the workspaces it can reach and its permissions. The assistant can only do what its permissions allow, and only in those workspaces.

Analytics and BrandKit aren't available over the public API. To reach them from another tool, use the MCP server. For technical details, see the MCP developer docs.

An AI assistant connected to Cevoid works with your analytics and your BrandKit. Each has its own permission: Analytics and BrandKit.

To answer a question, the assistant picks the metrics, breakdowns, filters, and date range, and runs the query in one workspace. It uses the same metrics, breakdowns, and filters as the report builder, with up to two breakdowns and an optional comparison with the previous period. Reports and dashboards it creates are saved in your workspace like the ones your team builds in the app.

An assistant's BrandKit changes go to your draft first. See BrandKit changes.

An AI assistant connected to Cevoid can't:

• Delete reports or dashboards
• Change default reports or default dashboards. It can run them and look them up.
• Upload logos, rename or delete custom fonts, or import your BrandKit from your website
• Read or change anything outside analytics and BrandKit, such as your reviews, posts, loyalty program members, profiles, flows, or emails
• Reach a workspace you didn't select when you connected it

You choose an AI assistant's permissions under Permissions on Cevoid's connect page, and can edit them later. Full access is selected by default.

Write includes Read. On Settings → AI & Agents → MCP⁠, an assistant's Permissions show Full access, Read only, or each custom permission: analytics:read, analytics:write, brandkit:read, or brandkit:write.

You connect an AI assistant with your own Cevoid account. It can reach only the workspaces where you're on the team.

1. Copy the Cevoid MCP server URL: https://mcp.cevoid.com/mcp
2. Paste it into your AI assistant, such as Claude, ChatGPT, or Cursor, as a custom MCP server or connector, and connect it there. A Cevoid page such as Connect Claude to Cevoid opens in your browser, after you log in if you aren't already.
3. Select Selected workspaces under Workspace access and select the workspaces in the list, or select All current workspaces
4. Select Full access, Read only, or Custom under Permissions. See Permissions.
5. Click Connect. Cevoid sends you back to your assistant to finish connecting. If your browser asks to open your assistant, allow it.

Selected workspaces starts with your current workspace. All current workspaces covers the workspaces you're on today: a workspace you join later needs a new connection.

Ask an AI assistant connected to Cevoid in plain language. Name the metric and the period, and the workspace if you connected several. For example:

• "Which posts had the most impressions in the last 30 days?"
• "Run the Top 24 Performing Posts report for last month."
• "Show post impressions by widget for the last quarter."
• "Create a report of post clicks by post label, and add it to a new dashboard."
• "Make the buttons in our BrandKit draft rounder, and show me a preview."
• "Publish our BrandKit draft."

Creating a report or a dashboard, and changing the BrandKit, need Write. See Permissions.

Date ranges such as the last 30 days run from the same time 30 days ago until now, not over whole days as in the app. Numbers can differ slightly from the same range in the app.

An AI assistant with BrandKit Write changes your BrandKit draft, the same draft your team edits in the BrandKit editor. Your store and emails keep the published BrandKit until the draft is published: ask the assistant to publish it, or publish it in the editor.

• Fonts: Fonts the assistant uploads, or imports from a public link, join your custom fonts right away. Using one in the draft is a draft change.
• Color schemes: Widgets and emails follow a renamed or deleted scheme when you publish, as with changes in the editor. The assistant can't delete your last scheme, or the scheme your popups or emails use.
• Icons: The assistant replaces icons with SVG code, picks an icon set, or returns icons to their defaults
• Previews: Preview images show sample content in your draft or published BrandKit, on desktop or mobile
• Website imports: While an import runs, the assistant can't change or publish the draft

The MCP server limits how much each person's assistant can do per minute.

Larger answers are refused. Narrow the question instead: a shorter date range, fewer metrics or breakdowns, or fewer rows.

An AI assistant's access to Cevoid lasts 30 days, and the assistant renews it automatically when you use it. Each renewal starts a new 30 days. The assistant's Expires date on Settings → AI & Agents → MCP⁠ shows when its access ends.

If an assistant goes 30 days without renewing, for example because you haven't used it, its status becomes Expired. Connect it again with the steps in Connect an AI assistant.

An assistant also loses access to a workspace when the team member who connected it leaves that workspace.

Settings → AI & Agents → MCP⁠ lists every AI assistant, or MCP client, that anyone on your team connected to your current workspace. Each one shows its Client, the Team member who connected it, its Status, and when it was Last used. Open one to see its details: the workspaces it reaches under Scope, such as Selected (2), its Permissions, and the date its access Expires.

Only the team member who connected an AI assistant can edit its permissions, while its status is Active. The change applies in every workspace the connection reaches.

1. Navigate to Settings → AI & Agents → MCP⁠
2. Open the assistant
3. Click Edit permissions
4. Select Full access, Read only, or Custom. With Custom, select None, Read, or Write for each resource.
5. Click Save changes

Removed permissions apply from the assistant's next request. The assistant gets what added permissions allow after it reconnects.

Permissions are the only part of a connection you can edit. To change the workspaces an assistant reaches, revoke it and connect it again.

Only the team member who connected an AI assistant, or an admin, can revoke it.

1. Navigate to Settings → AI & Agents → MCP⁠
2. Open the assistant
3. Click Revoke
4. Click Revoke to confirm. The assistant loses access right away. You can't undo this.

When you revoke an assistant you connected, it loses access to every workspace it reaches. When an admin revokes another team member's assistant, it loses access to the current workspace.

The connect page shows This connection request has expired when more than 10 minutes passed since your AI assistant started connecting, or when the request was already used, for example after you reloaded the page. Return to your assistant and connect it again.

An AI assistant can only do what its permissions allow, in the workspaces you connected. Check its Permissions and Scope on Settings → AI & Agents → MCP⁠, and edit its permissions to add one. The assistant gets what a new permission allows after it reconnects.