> For the complete documentation index, see [llms.txt](https://doc.commandersact.com/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://doc.commandersact.com/features/mcp-server-closed-beta.md). # MCP server (closed beta) ### Commanders Act MCP Server Setup Guide This guide explains how to connect Claude (or chatGPT, etc.) to the Commanders Act MCP server so your team can experiment with AI access to Commanders Act through the standard Model Context Protocol. ### 1. Overview The Commanders Act MCP server exposes Commanders Act platform capabilities as AI-callable tools through the standard Model Context Protocol. It is designed to be consumed by MCP compatible AI clients, without requiring any proprietary integration. The server uses OAuth 2.1 with PKCE for authentication and exposes a wide range of tools covering Sources, Destinations, Enrichments, Cleansing, Tag Management, Health, Reports and Admin. The MCP server is currently in **closed beta**. At this stage, only a limited number of users are testing it. Typical use cases include: * discovering the sites available to the authenticated user * exploring Sources, Destinations, Enrichments, Cleansing, Health, Reports, Tag Management, and Admin data * using Commanders Act data through AI workflows without building a proprietary integration ### 2. Access conditions Because the MCP server is currently in closed beta, access is not yet generally available. At this time, we are not accepting new users, but you can contact support team to be added to the waiting list. ### 3. Requirements To use the Commanders Act MCP server, you need: * a Commanders Act account * access to the MCP server as part of the closed beta * an MCP compatible AI client * a browser available for the authentication flow *** ### 4. Authentication The Commanders Act MCP server uses OAuth 2.1 with PKCE. The authentication flow is fully automatic and secure. No token is manually handled or exposed to the user. Typical flow: 1. The client attempts to call the MCP server 2. The server indicates that authentication is required 3. The client automatically discovers the authentication flow 4. A browser window opens with the Commanders Act login page 5. The user logs in with their Commanders Act credentials 6. The client securely completes the authentication process 7. The client includes the token in all subsequent tool calls Once authenticated, the connection is established and all interactions with the MCP server are seamless. *** ### 5. Connecting an MCP client The Commanders Act MCP server is exposed through a standard MCP interface. To connect a compatible AI client, you need to register the MCP server using the connection details provided by the Commanders Act support team. Depending on the client, this may involve: * adding the MCP server in a UI * or declaring it in a configuration file Once configured, the authentication flow will start automatically on first use. *** ### 6.a. Example client environment: Claude Code Claude Code is one example of an MCP compatible client. To connect the Commanders Act MCP server: 1. Open Claude 2. Go to Settings → Integrations 3. Enable the connector named “Commanders Act MCP” 4. When prompted, enter the MCP server URL provided by Commanders Act 5. Complete the authentication flow in the browser Once connected, you can start using the MCP server directly in conversations. To verify the connection, you can ask: “Can you list the available tools from the Commanders Act MCP?” You should see all available tools exposed by the server. ### 6.b. Example client environment: ChatGPT ChatGPT is another MCP compatible client. To connect the Commanders Act MCP server: 1. Open ChatGPT 2. Go to Settings → Apps & Connectors → Advanced 3. Enable Developer Mode 4. Go to Connectors and create a new connector 5. Use the MCP server connection details provided by Commanders Act 6. Select OAuth as the authentication method 7. Save the connector To use the MCP server in a conversation: 1. Start a new chat 2. Enable Developer Mode in the chat 3. Add the Commanders Act connector as a source Once connected, ChatGPT will be able to call MCP tools directly. To validate the connection, you can ask: Can you list the available tools from the Commanders Act MCP? ChatGPT will display tool calls in the interface. Some actions may require confirmation. *** ### 7. Available tool families The MCP server exposes more than 35 tools across the following areas: **Profile** * list accessible sites for the authenticated user **Sources** * list source categories * list configured sources * retrieve source details **Destinations** * list destination categories * list configured destinations * retrieve destination details **Enrichments and Cleansing** * list enrichments * retrieve enrichment details * list cleansing transformations * retrieve transformation details **Health and Monitoring** * retrieve source data quality * retrieve cookie scanners * retrieve scanner cookies * retrieve scanner install code * retrieve collection monitoring data * retrieve privacy statistics * retrieve health alerts * retrieve health notifications **Tag Management** * list web containers * retrieve container details * list web datalayer variables * list web triggers * list web constraints * list web perimeters * list server-side destinations * list server-side datalayer variables * list server-side tags **Explore and Reports** * list reports * retrieve report data **Data Management** * retrieve normalized datalayer **Admin** * retrieve activity logs * list users * list redirect rules *** ### 8. First validation After connecting and authenticating, you can validate the integration with a simple natural language request. Recommended prompt: Can you list the sites available to my Commanders Act account? You can also verify that the MCP server is correctly connected by asking: Can you list the available tools from the Commanders Act MCP? This confirms: * authentication is working * the MCP connection is active * the tools are correctly exposed * your permissions are correctly applied You can then continue with prompts such as: Can you list the configured sources for my account?\ Can you show me the configured destinations?\ Can you show me current health alerts?\ Can you list the available web containers?\ Can you list the available reports? *** ### 9. Practical prompt examples Once connected, you can ask Claude or ChatGPT questions about your actual Commanders Act environment, not just request technical lists. Replace \[site], country names, platforms and time ranges with your own. **Results depend on the account permissions and data exposed through the MCP tools.** Ask the assistant to distinguish verifiable findings from hypotheses. #### 9.1. Understand the marketing use cases in your account **Try this prompt:** > Analyze my Commanders Act configuration and identify the marketing use cases that appear to be implemented, such as server-side conversion tracking, audience activation, data enrichment and consent management. Explain the business value, data used and connected platforms for each. Present the result as a non-technical use-case catalog and distinguish verified configuration from assumptions. **What you can get:** A business-friendly overview of how your account is configured and which marketing activities it supports. #### 9.2. Investigate a sudden drop in conversions **Try this prompt:** > Conversions reported in Meta Ads have dropped since yesterday. For \[site], examine Commanders Act collection alerts, available event monitoring, source data quality and recent configuration activity. Build a timeline of relevant observations, list possible tracking-related explanations, and prioritize what I should check next. Separate evidence from hypotheses and state what you cannot verify without access to Meta Ads. **What you can get:** An evidence-based troubleshooting checklist and investigation timeline, without claiming to know the cause of external advertising results. #### 9.3. Discover opportunities in your CDP audiences **Try this prompt:** > Analyze my Commanders Act CDP segments. Group them by likely marketing objective: acquisition, retargeting, retention, reactivation, personalization or exclusion. Build a searchable, filterable table or interactive HTML page. Propose five relevant activation ideas using my existing destinations, and mark classifications inferred from segment names. **What you can get:** An organized, explorable segment catalog and new campaign ideas based on your current setup. ![Interactive CDP segments analysis in Claude with Commanders Act MCP](https://1259070148-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2F-Mk6XpTQ2LaRLcr2tA-d%2Fuploads%2FwiudB3hJFVeJWpy3TFJN%2Fmcp-claude-segments.webp?alt=media) #### 9.4. Visualize the journey of a purchase event **Try this prompt:** > For \[site], reconstruct how the purchase event is expected to move from collection through relevant variables, transformations, web or server-side tags and marketing destinations in Commanders Act. Draw a flow diagram. Clearly mark verified configuration relationships, inferred steps and anything the available MCP data cannot confirm. **What you can get:** A readable diagram of the event relationships visible in your configuration, useful for tracking documentation and investigations. ![Visualization of a purchase event configuration in Claude](https://1259070148-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2F-Mk6XpTQ2LaRLcr2tA-d%2Fuploads%2FDtlEDoGF9OYkPepi1BEx%2Fmcp-claude-purchase.webp?alt=media) #### 9.5. Compare tracking across countries or brands **Try this prompt:** > Compare the Commanders Act setups of my French, German and Italian sites. Which advertising destinations, web containers, server-side configurations and relevant tags are present on some sites but not others? Generate a site-by-site comparison matrix and highlight differences worth reviewing, without assuming each difference is an error. **What you can get:** A comparative matrix that makes multi-site governance and harmonization easier. ![Comparison of marketing configurations across sites in Claude](https://1259070148-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2F-Mk6XpTQ2LaRLcr2tA-d%2Fuploads%2FwYBNYeBZPmWPQ3Wo9eUI%2Fmcp-claude-multisite.webp?alt=media) #### 9.6. Prepare a cookie and privacy review **Try this prompt:** > Review the latest available cookie scanner results and privacy statistics for \[site]. Group detected cookies by vendor where possible, identify items requiring further investigation, and prepare a prioritized compliance review checklist. If possible, display the findings as a filterable dashboard. Do not claim a legal violation without sufficient supporting evidence. **What you can get:** A prioritized review based on observable cookie and privacy information. #### From answers to charts and interactive HTML pages Claude and ChatGPT can enrich their answers with **tables, charts, flow diagrams, custom dashboards and even interactive HTML pages**, depending on the AI client's features. The MCP provides the data; the AI client creates the visualizations. **Follow-up prompt:** > Turn the Commanders Act results from this conversation into an interactive HTML dashboard. Add a summary, relevant charts, a searchable table, filters and a detail panel. Use only retrieved data, and label assumptions or missing values clearly. You can refine the result in the same conversation: "Focus on reactivation audiences", "Add a chart by marketing objective" or "Summarize these insights for my team". **Closed beta limitations:** the current documented MCP capabilities primarily support reading and analyzing configurations, reports and monitoring information. They do not automatically publish tag changes or create campaigns. External platform performance metrics may not be available. Interactive HTML generation and export depend on the features supported by your chosen AI client. *** ### 10. Token lifetime Access tokens are time limited. The default lifetime is 8 hours. When the token expires, a new authentication is required. *** ### 11. Troubleshooting Login page does not open\ Make sure your client supports MCP with authentication and that a browser can be launched Authentication does not complete\ Verify that the MCP server connection details provided by support are correct and reachable Tools do not appear in Claude\ Make sure the “Commanders Act MCP” connector is enabled and authentication completed You can validate by asking: Can you list the available tools from the Commanders Act MCP? If no tools appear, try reconnecting the MCP server or restarting the session Connection works but no data is returned\ Start with a simple request such as listing available sites to verify permissions *** ### 12. Security and governance The Commanders Act MCP server relies on: * a standard MCP interface * OAuth based authentication * browser based login * secure token usage handled by the client This allows experimentation through approved MCP compatible AI clients without requiring a custom integration. As the server is currently in closed beta, access remains restricted. *** ### 13. Support For access or setup information, contact the Commanders Act support team. They can provide: * connection details * setup guidance * recommended first steps --- # Agent Instructions This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com. ## Querying This Documentation If you need additional information that is not directly available in this page, you can query the documentation by asking a question. Perform an HTTP GET request on the following URL with the `ask` and `goal` query parameters: ``` GET https://doc.commandersact.com/features/mcp-server-closed-beta.md?ask=&goal= ``` `ask` is the immediate question: it should be specific, self-contained, and written in natural language. `goal` is what the user is ultimately trying to achieve, the reason they need the answer. Sharing it helps GitBook give you a better, more relevant answer. A goal is most helpful when it describes the outcome the user wants rather than restating the question. For example, with `ask=how do I create an API token`, a goal like `automate deployments from our CI pipeline` lets GitBook tailor the answer to that use case. The response will contain a direct answer to the question and relevant excerpts and sources from the documentation. Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.