[Enhancement]: Support outbound proxy configuration for remote MCP HTTP/SSE transports

[inv-Eldho]

What happened?

I'm trying to connect a remote Splunk MCP server from LibreChat through a proxy, but I'm getting a connection timeout error.

The MCP endpoint is reachable from the API container using curl, so network connectivity appears to be working correctly. Since we are using a proxy, we added the proxy-related headers/configuration in the YAML request settings, and the connection is configured like this:

splunk:
    startup: false
    type: streamable-http
    url: <splunk_url>
    iconPath: /assets/splunk.png
    timeout: 80000
    initTimeout: 30000
    headers:
      Authorization: "Bearer {{splunk_api_key}}"
      Content-Type: "application/json"
      HTTPS_PROXY: <proxy_address>
      Accept: "application/json, text/event-stream"
    customUserVars:
      splunk_api_key:
        title: "Splunk API Key"
        description: "Enter your Splunk Bearer token"

We are using curl like this from the LibreChat container, and the connection is successful.

curl -v -X POST \
  --proxy <proxy_address>/ \
  -H "Authorization: Bearer <token>" \
  -H "Content-Type: application/json" \
  -H "Accept: application/json, text/event-stream" \
  -d '{"jsonrpc":"2.0","method":"initialize","params":{"protocolVersion":"2024-11-05","capabilities":{}},"id":1}' \
<splunk_url>

Could you please check this

Version Information

1bc2692

Steps to Reproduce

1. Connect the remote Splunk MCP server from LibreChat through a proxy.

What browsers are you seeing the problem on?

No response

Relevant log output

2026-05-07 21:06:13 info: [updateUserPluginsController] Attempting disconnect of MCP server "splunk" for user 6862786e99cd9d0127a61df8 after plugin auth update.
2026-05-07 21:06:13 info: [MCP Reinitialize] Reinitializing server: splunk
2026-05-07 21:06:13 info: [MCP Reinitialize] Disconnected existing user connection for server: splunk
2026-05-07 21:06:13 info: [MCP][User: 6862786e99cd9d0127a61df8][splunk] Establishing new connection
2026-05-07 21:06:14 info: [MCP][User: 6862786e99cd9d0127a61df8][splunk] Access token missing
2026-05-07 21:06:14 info: [MCP][User: 6862786e99cd9d0127a61df8][splunk] Access token missing and no refresh token available — re-authentication required
2026-05-07 21:06:14 info: [MCP][splunk][6862786e99cd9d0127a61df8] Re-authentication required for "splunk": access token missing and no refresh token available, will trigger OAuth flow
2026-05-07 21:06:14 info: [MCP][User: 6862786e99cd9d0127a61df8][splunk] Creating streamable-http transport: <url>
2026-05-07 21:06:24 error: [MCP][User: 6862786e99cd9d0127a61df8][splunk] Transport error (transient, will reconnect): fetch failed (request aborted, likely after a timeout — connection may still be usable)
2026-05-07 21:06:24 info: [MCP][User: 6862786e99cd9d0127a61df8][splunk] Streamable-http transport closed
2026-05-07 21:06:24 error: [MCP][User: 6862786e99cd9d0127a61df8][splunk] Connection failed: fetch failed
2026-05-07 21:06:26 info: [MCP][User: 6862786e99cd9d0127a61df8][splunk] Creating streamable-http transport: <url>
2026-05-07 21:06:37 error: [MCP][User: 6862786e99cd9d0127a61df8][splunk] Transport error (transient, will reconnect): fetch failed (request aborted, likely after a timeout — connection may still be usable)
2026-05-07 21:06:37 info: [MCP][User: 6862786e99cd9d0127a61df8][splunk] Streamable-http transport closed
2026-05-07 21:06:37 error: [MCP][User: 6862786e99cd9d0127a61df8][splunk] Connection failed: fetch failed
2026-05-07 21:06:41 info: [MCP][User: 6862786e99cd9d0127a61df8][splunk] Creating streamable-http transport: <url>
2026-05-07 21:06:44 error: [MCP][User: 6862786e99cd9d0127a61df8][splunk] Failed to establish connection Connection timeout after 30000ms
2026-05-07 21:06:44 info: [MCP Reinitialize] getConnection threw error: Connection timeout after 30000ms
2026-05-07 21:06:44 info: [MCP Reinitialize] OAuth state - oauthRequired: false, oauthUrl: null
2026-05-07 21:06:44 error: [MCP Reinitialize] Error initializing MCP server splunk for user: Connection timeout after 30000ms

Screenshots

None

Code of Conduct

• I agree to follow this project's Code of Conduct

[danny-avila]

Thanks for the detailed report.

I checked the MCP streamable-http connection path, and the issue appears to be proxy handling rather than Splunk reachability/auth.

In the current config, HTTPS_PROXY is being added under headers, but that is not equivalent to curl --proxy. LibreChat will send it as a normal HTTP request header to the MCP server; it will not configure the outbound network proxy.

For MCP streamable-http, LibreChat currently creates an undici transport/agent directly and does not attach a proxy dispatcher from MCP YAML headers. That explains why the same endpoint works with:

curl --proxy <proxy_address> ...

but times out from LibreChat.

For now, please remove HTTPS_PROXY from the MCP headers block. A valid workaround would be to configure proxying at the container/network layer so outbound traffic from the API container is routed through your proxy.

I’ll track this as a LibreChat feature gap: MCP remote HTTP/SSE transports should support outbound proxy configuration, likely via a proper top-level MCP proxy setting or by honoring LibreChat’s existing PROXY env var in the MCP transport path.

[inv-Eldho]

@danny-avila
Thank you for the immediate reply and workaround. I added the proxy settings at the container level and tried again, but it’s still not connecting to the Splunk MCP server. I added the proxy configuration in the LibreChat container as environment variables. I also tried adding the PROXY key in the .env file, but I’m still getting the same error.

- HTTPS_PROXY=<proxy_url>
- HTTP_PROXY= <proxy_url>
- NO_PROXY=localhost,127.0.0.1,::1,host.docker.internal

Logs :

2026-05-07 22:23:09 error: [MCP][User: 6862786e99cd9d0127a61df8][splunk] Transport error (transient, will reconnect): fetch failed (request aborted, likely after a timeout — connection may still be usable)
2026-05-07 22:23:09 info: [MCP][User: 6862786e99cd9d0127a61df8][splunk] Streamable-http transport closed
2026-05-07 22:23:09 error: [MCP][User: 6862786e99cd9d0127a61df8][splunk] Connection failed: fetch failed
2026-05-07 22:23:11 info: [MCP][User: 6862786e99cd9d0127a61df8][splunk] Creating streamable-http transport: <splunk_url>
2026-05-07 22:23:21 error: [MCP][User: 6862786e99cd9d0127a61df8][splunk] Transport error (transient, will reconnect): fetch failed (request aborted, likely after a timeout — connection may still be usable)
2026-05-07 22:23:21 info: [MCP][User: 6862786e99cd9d0127a61df8][splunk] Streamable-http transport closed
2026-05-07 22:23:21 error: [MCP][User: 6862786e99cd9d0127a61df8][splunk] Connection failed: fetch failed
2026-05-07 22:23:25 info: [MCP][User: 6862786e99cd9d0127a61df8][splunk] Creating streamable-http transport: <splunk_url>
2026-05-07 22:23:28 error: [MCP][User: 6862786e99cd9d0127a61df8][splunk] Failed to establish connection Connection timeout after 30000ms
2026-05-07 22:23:28 info: [MCP Reinitialize] getConnection threw error: Connection timeout after 30000ms
2026-05-07 22:23:28 info: [MCP Reinitialize] OAuth state - oauthRequired: false, oauthUrl: null
2026-05-07 22:23:28 error: [MCP Reinitialize] Error initializing MCP server splunk for user: Connection timeout after 30000ms
2026-05-07 22:23:36 error: [MCP][User: 6862786e99cd9d0127a61df8][splunk] Transport error (transient, will reconnect): fetch failed (request aborted, likely after a timeout — connection may still be usable)
2026-05-07 22:23:36 info: [MCP][User: 6862786e99cd9d0127a61df8][splunk] Streamable-http transport closed
2026-05-07 22:23:36 warn: [MCP][User: 6862786e99cd9d0127a61df8][splunk] Circuit breaker: too many failures, backing off for 30000ms
2026-05-07 22:23:36 error: [MCP][User: 6862786e99cd9d0127a61df8][splunk] Connection failed: fetch failed
2026-05-07 22:23:36 error: [MCP][splunk][6862786e99cd9d0127a61df8] Failed to connect after 3 attempts fetch failed

[danny-avila]

Thanks for testing. This is expected with the current MCP transport: HTTP_PROXY, HTTPS_PROXY, and LibreChat’s PROXY env var are not currently applied to MCP streamable-http requests.

Those env vars only help if the HTTP client reads them, and this MCP path uses an explicit undici dispatcher. So this issue remains scoped to adding outbound proxy support for remote MCP HTTP/SSE transports.

[inv-Eldho]

@danny-avila
Thank you for the update. I’m waiting for this update to be merged into the main branch

[danny-avila]

Closed by #13076