Problem
When an MCP server fails to connect or reconnect (via /mcp), the only feedback is:
Failed to reconnect to <server-name>.
There is no additional information about why it failed — no stderr output, no exit code, no timeout indication, nothing. This makes debugging MCP server startup issues extremely difficult.
Expected behavior
The error message should include at least:
- The command that was attempted
- stderr output from the failed process (if any)
- Whether it was a timeout vs crash vs connection refused
- The exit code if the process exited
For example:
Failed to reconnect to hn.
Command: bash -c "exec clojure -M -m hn.mcp"
Exit code: 1
stderr: Exception in thread "main" java.io.FileNotFoundException: ...
Context
I'm building custom MCP servers (Clojure stdio servers). When they fail to start, the only way to debug is to manually run the command in a terminal and see what happens. The /mcp dialog gives zero diagnostic info.
claude mcp get <name> sometimes reports ✓ Connected even when /mcp in-session fails to reconnect, which adds to the confusion.
Environment
- Claude Code on Linux (WSL2)
- MCP servers configured via
claude mcp add-json --scope user
- stdio transport
Problem
When an MCP server fails to connect or reconnect (via
/mcp), the only feedback is:There is no additional information about why it failed — no stderr output, no exit code, no timeout indication, nothing. This makes debugging MCP server startup issues extremely difficult.
Expected behavior
The error message should include at least:
For example:
Context
I'm building custom MCP servers (Clojure stdio servers). When they fail to start, the only way to debug is to manually run the command in a terminal and see what happens. The
/mcpdialog gives zero diagnostic info.claude mcp get <name>sometimes reports✓ Connectedeven when/mcpin-session fails to reconnect, which adds to the confusion.Environment
claude mcp add-json --scope user