fix: reconnect stale SSE MCP sessions

[mturac]

Description

Closes #12431.

This makes SSE MCP tool, prompt, and resource calls recover from stale session or closed transport errors by resetting the client/transport, reconnecting, and retrying the operation once. The reconnect path creates a fresh MCP client so restarted SSE servers do not receive requests with an old session.

Checklist

• I've read the contributing guide
• The relevant docs, if any, have been updated or created
• The relevant tests, if any, have been updated or created

Screen recording or screenshot

Not applicable; this is MCP connection recovery behavior.

Tests

• npm run vitest -- context/mcp/MCPConnection.vitest.ts
• npm run tsc:check
• git diff --check

---

Summary by cubic

Fixes SSE MCP calls that fail after stale or closed sessions by resetting the client/transport, reconnecting, and retrying once. Improves reliability when SSE servers restart or drop connections.

• Bug Fixes
  • Added an SSE-only reconnect-and-retry wrapper that detects session/transport-closed errors, resets the Client/Transport, calls connectClient(true, ...), then retries once.
  • Wrapped getResource, getPrompt, and callTool; updated MCPManagerSingleton and core/tools/callTool.ts to use these connection methods.
  • Added tests for SSE tool-call retry and to ensure non-SSE calls do not retry.

^{Written for commit e57c06f. Summary will update on new commits. Review in cubic}

[cubic-dev-ai]

No issues found across 4 files

_{Re-trigger cubic}