How do I verify Montage workspace access with Mastra?
Mastra separates MCPClient, which consumes remote tools, from MCPServer, which exposes agents and workflows to other systems. Its current documentation also distinguishes direct client configuration from platform-managed connections. Montage fits the consuming-client path here: use a stable named server, perform supported authorization, and inspect one real status response before exposing a larger tool collection to an agent.
Using Mastra, call only Montage signals_status as the first live operation. Report the actual workspace identity and coverage from the tool response. If authentication, transport, or the tool is unavailable, stop and explain the blocker. Do not start collection, recurring tracking, writes, publishing, messaging, or social actions.Choose the consuming client API
Match the installed @mastra/mcp version to the current MCP guide and migration notes. Instantiate MCPClient rather than MCPServer for a hosted Montage connection. Platform Connect can manage supported hosted integrations, but this guide does not claim Montage has a managed catalog entry. Keep the direct client's identity and lifecycle explicit within the application's existing server-side structure.
Configure the named remote server
Under servers, add montage with url as new URL('https://app.montages.co/mcp'). Keep its client id stable and preserve other configured servers. Local command and args examples apply to subprocess servers, not Montage. Review the current remote transport and protocol compatibility before assuming constructor success proves the application reached or authorized the service.
Use the supported authorization flow
For OAuth-protected servers, current Mastra documentation points to authenticate() and its OAuth configuration guide. Implement that flow deliberately, with secure per-user credential storage, or use an existing approved header credential through documented requestInit settings. Do not commit secrets in the client module. A Mastra agent's model-provider account does not establish Montage workspace access.
Inspect named tools before agent attachment
Call listTools() and locate the actual status tool, accounting for server-name prefixes. Do not hand the complete returned object to an agent simply because the example does so. Generated schema types can improve development ergonomics but are optional and do not prove authorization. Handle missing tools and validation errors explicitly rather than asserting every expected execute function exists.
Execute status and disconnect
Make signals_status the first live Montage operation and narrow its returned value before reading workspace fields. The documentation warns that execution may return an envelope, validation error, or undefined instead of assumed structured output. Disconnect in cleanup after testing. No status success grants permission for automatic collection, workflows, publishing, or social actions through the agent.
Keep the boundaries clear.
- Current @mastra/mcp APIs differ from 1.x; direct MCPClient configuration is not a claim of a managed Montage platform connection.
- Documented setup, not live interoperability certification. A visible tool catalog is not workspace authorization.
- First live operation: read-only signals_status. No automatic collection, recurring tracking, or social actions; broader actions need separate explicit authorization.
Start with a read-only check: ask your agent to report actual source coverage before collecting records or exporting an audience.
Primary references
Documentation reviewed 2026-11-15. Product capabilities, client availability and plans can change. Third-party marks identify their products, not an affiliation.