mirror of
https://github.com/czlonkowski/n8n-mcp.git
synced 2026-04-03 16:13:08 +00:00
12d7d5bdb6
Fix SSE clients entering rapid reconnection loops because POST /mcp never routed messages to SSEServerTransport.handlePostMessage() (#617). Root cause: SSE sessions were stored in a separate `this.session` property invisible to the StreamableHTTP POST handler. The POST handler only checked `this.transports` (StreamableHTTP map), so SSE messages were never delivered, causing immediate reconnection and rate limiter exhaustion. Changes: - Add GET /sse + POST /messages endpoints following the official MCP SDK backward-compatible server pattern (separate endpoints per transport) - Store SSE transports in the shared this.transports map with instanceof guards for type discrimination - Remove legacy this.session singleton, resetSessionSSE(), and isExpired() - Extract duplicated auth logic into authenticateRequest() method - Add Bearer token auth and rate limiting to SSE endpoints - Add skipSuccessfulRequests to authLimiter to prevent 429 storms - Mark SSE transport as deprecated (removed in MCP SDK v2.x) The handleRequest() codepath used by the downstream SaaS backend (N8NMCPEngine.processRequest()) is unchanged. Session persistence (exportSessionState/restoreSessionState) is unchanged. Closes #617 Conceived by Romuald Członkowski - https://www.aiadvisors.pl/en Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>