Anaplan AI Gateway connects external AI systems to Anaplan through a governed Model Context Protocol (MCP) interface. Use it to connect your Anaplan models to AI systems such as Gemini Enterprise and Claude. Access is read-only and based on the user’s Anaplan credentials. Query Anaplan model data safely knowing that the calculations are correct, and that users can only see what they are allowed to see.

Note: AI Gateway is an entitled service that requires a subscription. Your Customer Success Business Partner can help you with this.

Note: Calculations derived from an Anaplan model and returned through AI Gateway will be correct. How your external AI system interprets these calculations is outside Anaplan control. Also, external LLMs have no inherent knowledge of model logic or semantics. Anaplan Analyst has greater drill-down capabilities and offers an enhanced and fully auditable experience grounded in Anaplan planning fundamentals. Learn more at Anaplan Analyst .

In Agent Studio, AI administrator role users enable AI Gateway by workspace. Users can query Anaplan models in enabled workspaces, based on their existing access permissions. New workspaces are off by default. Learn more at Enable Anaplan AI Gateway.

Note: Before enabling AI Gateway, review your model access control and ensure sensitive data is adequately restricted. Review roles, model access, selective access, and Dynamic Cell Access (DCA). Also note that UI based access control such as filters won’t stop visibility through AI Gateway. The readiness advice for model builders in this Community article for Anaplan Analyst also applies to AI Gateway.

AI Gateway is a remote MCP server. Your chosen AI system must support remote MCP servers. Connection is established through an AI Gateway URL and a client ID for your AI system in your region. Connect external AI to Anaplan AI Gateway lists the AI Gateway URL and region client IDs. It also has information on creating connectors and MCP client requirements.

Note: Your IT department may have already configured MCP connectors for the AI system used by your organization. Check with your IT department.

Irrespective of the connector used, when users connect to Anaplan through AI Gateway, they are prompted for their Anaplan user credentials and authenticated as normal.

Subject to the user’s access permissions, AI systems connected through AI Gateway can see all models in the workspace. It’s important to direct prompts to the correct model. As a starting point, you can ask what workspaces and models you can see. You can then bind the session to one model with the prompt:

Set the model context to model_name.

Note: If you don't direct your AI system to a specific model, it generally tries to find the model most relevant to your prompt and binds to it.

AI Gateway limits access to three models per hour per user. You can switch between the three models as often as required.

When using an Anaplan connector, we recommend that you disconnect the other connectors in your MCP client. This helps limit your session’s scope to Anaplan data.

Note: If your session doesn’t return workspaces or models, check with your Anaplan AI administrator to ensure that the workspace is enabled for AI Gateway. Also, check that you have access to the workspace and models.