How do I verify Montage workspace access with LibreChat?
LibreChat can attach MCP servers to a normal conversation or select their tools in Agent Builder. Those are distinct configuration paths, and the normal composer picker is hidden for Agents and Assistants endpoints. Montage setup should therefore start with the actual conversation type, not a missing-menu diagnosis. Per-user connection state and a verified workspace response matter more than whether another user sees the same server name.
Using LibreChat, 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 chat or Agent Builder
For a regular chat, open the composer's Attach and tools palette and inspect MCP Servers. For an agent, use Agent Builder, Tools, Add, then the MCP server filter in the Tool Library. Do not expect the regular chat picker on Agents or Assistants endpoints. Pick a model that supports tool calls before interpreting a silent model response as a failure of remote connectivity.
Locate or register the intended server
Inspect MCP Settings and the deployment's supported custom-server configuration before selecting Montage. Use https://app.montages.co/mcp as the remote address, not a local shell command. An administrator may control which servers appear in chat, so absence can be an availability policy rather than a network problem. Avoid changing a deployment-wide configuration when only one user's conversation attachment or authorization needs repair.
Use the per-user sign-in path
LibreChat documents Needs Auth for OAuth servers and opens sign-in when a selected server requires it. Complete that flow for the intended Montage account. Its documentation distinguishes per-user connections from another person's tokens or saved credentials; do not try to reuse a colleague's successful connection. If your deployment cannot support the required remote flow, report it as blocked rather than inventing a consent URL.
Narrow the attached tools
In Agent Builder, keep only the desired status tool instead of using Select all. In regular chat, attaching a server can expose every tool it publishes, so keep the request explicit and preserve approval controls. Connected and On-demand describe connection behavior, not workspace permission. Likewise, a catalog changing without a restart does not grant standing authorization for newly added mutations or social operations.
Verify the response in the same conversation
Call Montage signals_status as the first live operation and inspect the returned workspace and coverage. If the server reports Error or Needs Auth, fix that specific condition before research; do not delete and re-add it reflexively. Keep the actual tool response separate from the model's summary. No successful conversation attachment authorizes automatic collection, recurring tracking, publishing, or other social actions on your behalf.
Keep the boundaries clear.
- Normal chat, Agents, and Assistants use different tool-selection interfaces; authorization remains per user.
- 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.