Connect with MCP

BETA

Agent Gateway is in BETA, and we're continuing to expand its capabilities. Your feedback helps shape what we build next—reach out to us through Relewise Support.

MCP lets an AI assistant discover and call Agent Gateway tools through the Model Context Protocol.

Before connecting, ensure that you have Use Agent Gateway permission and that MCP is enabled for the Dataset. See Security and Permissions and Configuration.

On the Agent Gateway page, select Connection Guide, then open the MCP tab. Copy the AI Tool Instructions directly into the AI tool, or use the Manual Setup values to configure the connection yourself.

Configure the Connection

One MCP connection can work with multiple Datasets. Use this endpoint:

https://my.relewise.com/agents/mcp

Use relewise-agent-gateway as the connection name.

Authenticate with OAuth

OAuth is recommended when the client supports interactive authorization. The client should discover the authorization configuration from the MCP endpoint. See OAuth for the supported registration methods, scopes, and authorization flow.

Authenticate with a Personal Access Token

Otherwise, configure the client's secure credential mechanism to send a Personal Access Token with every request:

Authorization: Bearer <personal-access-token>

Keep the token in the client's credential settings; do not paste it into chat.

Verify Identity and Select a Dataset

Call get_me with no arguments. It returns:

• user: displayName and emailAddress.
• authentication: non-secret metadata with method (OAuth or PersonalAccessToken) and displayName. Personal Access Token metadata also includes nullable description and expiresAt. OAuth metadata includes clientId, connectionKey, sessionKey, accessTokenExpiresAt, scopes, and resource. No token secret is returned.
• licenses: License references with licenseId, displayName, and datasets. Each Dataset has datasetId, displayName, and type.

An empty licenses collection means no Datasets are available for your user and the Connected App's Dataset access or Personal Access Token's Dataset Scope.

After choosing a Dataset, pass its datasetId GUID with each tool call that works with that Dataset. If the choice is ambiguous, ask the user rather than guessing. Call get_dataset_details with that ID to check its Agent Gateway settings.

Available tools may require permissions or capabilities that are not enabled for your selected Dataset.

Browse Areas for available tools. Use Dataset metadata to find valid languages, currencies, and Data Keys when needed. For connection or access issues, see Troubleshooting and FAQ.