MCP Streamable HTTP transport: AbortError on reconnection causes intermittent "Connection closed" failures

[estebanrestoyagicap]

What happened?

Title: MCP Streamable HTTP transport: AbortError on reconnection causes intermittent "Connection closed" failures

Description:

When connecting LibreChat to an MCP server using Streamable HTTP transport (hosted on n8n), the connection works for approximately 15 tool calls, then crashes with:

Error processing tool: [MCP][server-name][tool_name] tool call failed: MCP error -32000: Connection closed

The error appears seemingly random (~1 in 15 calls).

Environment

• LibreChat (latest)
• MCP server: n8n (Streamable HTTP + SSE transports)
• Tested with both librechat.yaml config and UI configuration

What I tested

Test	Result
LibreChat → MCP server (Streamable HTTP)	❌ Crashes after ~15 requests
LibreChat → MCP server (SSE)	❌ Same issue (see finding #2 below)
MCP Inspector (npx @modelcontextprotocol/inspector) → same server	✅ Works perfectly
Manual curl POST initialize (Streamable HTTP)	✅ Correct response
Manual curl GET SSE stream (120s)	✅ Connection maintained, no timeout

The MCP server works correctly. The issue is on the LibreChat client side.

Version Information

Version :

Configuration version (required)

version: 1.3.1

LibreChat v0.8.2-rc1

Steps to Reproduce

1. Add a mcp server
2. Try to call a tool multiple time in a row
3. get randomly disconnected

What browsers are you seeing the problem on?

Chrome

Relevant log output

{"event":{},"level":"error","message":"[MCP][pkm-knowledge] Transport error: SSE error: undefined","stack":"Error: SSE error: undefined\n    at _eventSource.onerror (/app/node_modules/@modelcontextprotocol/sdk/dist/cjs/client/sse.js:101:31)\n    at EventSource.scheduleReconnect_fn (/app/node_modules/eventsource/dist/index.cjs:194:53)\n    at /app/node_modules/even...","timestamp":"2026-02-12T08:11:24.222Z"}
{"level":"info","message":"[MCP][pkm-knowledge] Reconnecting 1/3 (delay: 2000ms)","timestamp":"2026-02-12T08:11:24.223Z"}
{"event":{},"level":"error","message":"[MCP][pkm-knowledge] Transport error: SSE error: undefined","stack":"Error: SSE error: undefined\n    at _eventSource.onerror (/app/node_modules/@modelcontextprotocol/sdk/dist/cjs/client/sse.js:101:31)\n    at EventSource.scheduleReconnect_fn (/app/node_modules/eventsource/dist/index.cjs:194:53)\n    at /app/node_modules/even...","timestamp":"2026-02-12T08:11:25.220Z"}
{"level":"info","message":"[MCP][pkm-knowledge] Reconnecting 1/3 (delay: 2000ms)","timestamp":"2026-02-12T08:11:25.220Z"}
{"event":{},"level":"error","message":"[MCP][pkm-knowledge] Transport error: SSE error: undefined","stack":"Error: SSE error: undefined\n    at _eventSource.onerror (/app/node_modules/@modelcontextprotocol/sdk/dist/cjs/client/sse.js:101:31)\n    at EventSource.scheduleReconnect_fn (/app/node_modules/eventsource/dist/index.cjs:194:53)\n    at /app/node_modules/even...","timestamp":"2026-02-12T08:11:25.720Z"}
{"level":"info","message":"[MCP][pkm-knowledge] Reconnecting 1/3 (delay: 2000ms)","timestamp":"2026-02-12T08:11:25.720Z"}
{"level":"info","message":"[MCP][pkm-knowledge] SSE transport closed","timestamp":"2026-02-12T08:11:26.224Z"}
{"level":"info","message":"[MCP][pkm-knowledge] Creating SSE transport: *****","timestamp":"2026-02-12T08:11:26.224Z"}
{"level":"info","message":"[MCP][pkm-knowledge] SSE transport closed","timestamp":"2026-02-12T08:11:27.222Z"}
{"level":"info","message":"[MCP][pkm-knowledge] Creating SSE transport: ****","timestamp":"2026-02-12T08:11:27.222Z"}
{"level":"info","message":"[MCP][pkm-knowledge] SSE transport closed","timestamp":"2026-02-12T08:11:27.721Z"}
{"level":"info","message":"[MCP][pkm-knowledge] Creating SSE transport: ****","timestamp":"2026-02-12T08:11:27.721Z"}
{"event":{},"level":"error","message":"[MCP][pkm-knowledge] Transport error: SSE error: undefined","stack":"Error: SSE error: undefined\n    at _eventSource.onerror (/app/node_modules/@modelcontextprotocol/sdk/dist/cjs/client/sse.js:101:31)\n    at EventSource.scheduleReconnect_fn (/app/node_modules/eventsource/dist/index.cjs:194:53)\n    at /app/node_modules/even...","timestamp":"2026-02-12T08:11:28.715Z"}
{"level":"info","message":"[MCP][pkm-knowledge] Reconnecting 1/3 (delay: 2000ms)","timestamp":"2026-02-12T08:11:28.715Z"}
{"level":"info","message":"[MCP][pkm-knowledge] SSE transport closed","timestamp":"2026-02-12T08:11:30.717Z"}
{"level":"info","message":"[MCP][pkm-knowledge] Creating SSE transport: ******","timestamp":"2026-02-12T08:11:30.717Z"}

Screenshots

Code of Conduct

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

[danny-avila]

This issue cannot be reproduced and seems specific to the MCP server, which wasn’t shared.

See: #9846

you should share as much of a reproduction as possible from the n8n MCP server