MCPRemoteProxy in Kubernetes fails to Initialize when transport `!= stdio`

[snorlaX-sleeps]

Bug description

The MCPRemoteProxy runs the proxyrunner container to proxy connections to a upstream MCP server.
The proxyrunner container will always fail to initialise as it is trying to connect to localhost, it does not attempt to connect to the upstream remoteURL but will use the configured port.

Flow:

• localhost is hardcoded here
• Generates the URL to connect to here
• then waitForInitializeSuccess runs for 5mins
• and finally will always error out but continue here

The remoteURL is setup and configured here but the value is then not used or overwritten to generate the serverURL.

Steps to reproduce

Deploy a MCPRemoteProxy with streamable-http in Kubernetes using the Operator (mine was using ghcr.io/stacklok/toolhive/proxyrunner:v0.12.2)
e.g Deploy a Github MCPRemoteProxy

Expected behavior

When the container is initialising it should be initialising with the upstream MCP server

Actual behavior

When the container is initialising it is hardcoded to contact localhost

Environment (if relevant)

• ToolHive version: v0.12.2 of the Operator and associated CRDs

Additional context

Unsure if this intended functionality but it will always error out after 5mins. This does not affect the MCPRemoteProxy functionality, as far as can be told.

Logs:

{"time":"2026-03-19T18:37:49Z","level":"INFO","msg":"Successfully loaded configuration from /etc/runconfig/runconfig.json"}
{"time":"2026-03-19T18:37:49Z","level":"INFO","msg":"auto-discovered and loaded configuration from runconfig.json file"}
{"time":"2026-03-19T18:37:49Z","level":"WARN","msg":"Secret manager not available, OAuth tokens will not be persisted","error":"secrets provider not configured. Please run 'thv secret setup' to configure a secrets provider first"}
{"time":"2026-03-19T18:37:49Z","level":"INFO","msg":"Waiting for MCP server to be ready","endpoint":"http://localhost:8081/mcp","timeout":300000000000}
...
{"time":"2026-03-19T18:42:50Z","level":"WARN","msg":"initialize not successful, but continuing","error":"initialize not successful after 5m1.300990965s (155 attempts)"}
Workload started successfully. Press Ctrl+C to stop.

[snorlaX-sleeps]

After enabling the debug logs it will then appear to skip any healthchecks on the upstream as the initialise failed:

{"time":"2026-03-20T15:06:13Z","level":"DEBUG","msg":"setting up transport","transport":"streamable-http"}
{"time":"2026-03-20T15:06:13Z","level":"WARN","msg":"Secret manager not available, OAuth tokens will not be persisted","error":"secrets provider not configured. Please run 'thv secret setup' to configure a secrets provider first"}
{"time":"2026-03-20T15:06:13Z","level":"DEBUG","msg":"Using bearer token authentication"}
{"time":"2026-03-20T15:06:13Z","level":"DEBUG","msg":"starting transport","transport":"streamable-http","container":"upstream-proxy"}
{"time":"2026-03-20T15:06:13Z","level":"DEBUG","msg":"setting up transparent proxy to forward to remote URL","port":8081,"target":"https://upstreamURL","base_path":"/mcp","raw_query":""}
{"time":"2026-03-20T15:06:13Z","level":"DEBUG","msg":"applied middleware","name":"oauth-token-injection"}
{"time":"2026-03-20T15:06:13Z","level":"DEBUG","msg":"applied middleware","name":"recovery"}
{"time":"2026-03-20T15:06:13Z","level":"DEBUG","msg":"applied middleware","name":"audit"}
{"time":"2026-03-20T15:06:13Z","level":"DEBUG","msg":"applied middleware","name":"usagemetrics"}
{"time":"2026-03-20T15:06:13Z","level":"DEBUG","msg":"applied middleware","name":"mcp-parser"}
{"time":"2026-03-20T15:06:13Z","level":"DEBUG","msg":"applied middleware","name":"auth"}
{"time":"2026-03-20T15:06:13Z","level":"DEBUG","msg":"rfc 9728 OAuth discovery endpoint enabled at /.well-known/oauth-protected-resource"}
{"time":"2026-03-20T15:06:13Z","level":"DEBUG","msg":"http transport started","container":"","port":8081}
{"time":"2026-03-20T15:06:13Z","level":"DEBUG","msg":"mcp server started successfully","container":"upstream-proxy"}
{"time":"2026-03-20T15:06:13Z","level":"INFO","msg":"Waiting for MCP server to be ready","endpoint":"http://localhost:8081/mcp","timeout":300000000000}
{"time":"2026-03-20T15:06:13Z","level":"DEBUG","msg":"Server returned status","status_code":401,"attempt":1}
{"time":"2026-03-20T15:06:13Z","level":"DEBUG","msg":"Server returned status","status_code":401,"attempt":2}
{"time":"2026-03-20T15:06:13Z","level":"DEBUG","msg":"Server returned status","status_code":401,"attempt":3}
{"time":"2026-03-20T15:06:14Z","level":"DEBUG","msg":"Server returned status","status_code":401,"attempt":4}
{"time":"2026-03-20T15:06:14Z","level":"DEBUG","msg":"Server returned status","status_code":401,"attempt":5}
{"time":"2026-03-20T15:06:16Z","level":"DEBUG","msg":"Server returned status","status_code":401,"attempt":6}
{"time":"2026-03-20T15:06:18Z","level":"DEBUG","msg":"Server returned status","status_code":401,"attempt":7}
{"time":"2026-03-20T15:06:20Z","level":"DEBUG","msg":"Server returned status","status_code":401,"attempt":8}
{"time":"2026-03-20T15:06:22Z","level":"DEBUG","msg":"Server returned status","status_code":401,"attempt":9}
{"time":"2026-03-20T15:06:23Z","level":"DEBUG","msg":"mcp server not initialized yet, skipping health check","target":"https://upstreamURL"}
{"time":"2026-03-20T15:06:24Z","level":"DEBUG","msg":"Server returned status","status_code":401,"attempt":10}

...

{"time":"2026-03-20T15:11:13Z","level":"DEBUG","msg":"mcp server not initialized yet, skipping health check","target":"https://upstreamURL"}
{"time":"2026-03-20T15:11:14Z","level":"DEBUG","msg":"Server returned status","status_code":401,"attempt":155}
{"time":"2026-03-20T15:11:14Z","level":"WARN","msg":"initialize not successful, but continuing","error":"initialize not successful after 5m1.320945768s (155 attempts)"}
{"time":"2026-03-20T15:11:14Z","level":"DEBUG","msg":"server has no group, updating globally registered clients for backward compatibility","server":"upstream-proxy","count":0}
{"time":"2026-03-20T15:11:14Z","level":"DEBUG","msg":"no target clients found for server","server":"upstream-proxy"}
Workload started successfully. Press Ctrl+C to stop.
{"time":"2026-03-20T15:11:23Z","level":"DEBUG","msg":"mcp server not initialized yet, skipping health check","target":"https://upstreamURL"}
{"time":"2026-03-20T15:11:33Z","level":"DEBUG","msg":"mcp server not initialized yet, skipping health check","target":"https://upstreamURL"}

[lujunsan]

Hey @snorlaX-sleeps ! Sorry for the delay getting back to you on this issue.
I have attempted to replicate the issue on the latest operation version, and wasn't able to; I might have missed something on how the issue works however. Is this still an ongoing issue for you? Have you tried it on the latest versions?
Let me know and I'll make sure we take a deeper look at this if it is still occurring.

[snorlaX-sleeps]

Thanks for getting back to me @lujunsan
There's been a major rewrite to the code base so it may have been resolved 🙂
I will be deploying more MCPRemoteProxies towards the end of the month, if I do not see it again I will close this out

However localhost is still hardcoded here before waiting for initialisation