Limitations
Archive-only data source
The MCP server currently has these limitations:
MCP runs queries only against archive storage. It does not support environments that use only high-tier storage (Frequent Search).
This can lead to the following scenarios:
Time-based log queries, such as show logs in last 10 minutes, return zero results even though you see logs in Frequent Search.
Any natural language query that targets recent logs appears empty because MCP cannot access non-archived data.
To avoid these issues, configure archive storage for your account and ensure relevant logs are being archived. For setup steps, see Connect S3 archive.
Single active connection per MCP server
The MCP server supports only one effective active connection per server at a time when multiple clients use the same server and tool set.
When you connect two tools or two workspaces to the same MCP server in parallel (for example, two Cursor connections pointing at the same MCP server), tool calls are routed to the first active connection because the server exposes tools with the same names. The second connection might:
Not receive tool responses
Appear to “hang” or behave inconsistently when invoking tools
Show results that seem to belong to another session
To work around this limitation:
Keep only one connection to the MCP server active at a time.
Disable or disconnect the MCP server in one client or workspace before enabling it in another.