[BUG] MCP server manual selection UI missing in v0.4.3 - regression

Summary

When multiple MCP servers expose tools with identical names (e.g., two apollo-mcp-server instances connecting to different Apollo Portals), there is no way to manually select which MCP server the agent should use. The agent can only guess based on conversation context, which is unreliable.

Reproduction Steps

  1. Configure two MCP servers using the same apollo-mcp-server command but different --url and --token args (connecting to different Apollo Portal instances)
  2. Both servers expose the same 17 tools (e.g., get_item, list_namespaces, publish_release)
  3. In chat, ask the agent to query config from one specific Apollo Portal
  4. The agent has no way to know which MCP server to target — there is no UI picker or explicit selection mechanism

Expected Behavior

In a previous version, users could explicitly select which MCP server to use before the agent invokes a tool. This selection UI allowed disambiguation when multiple servers expose identically-named tools.

Actual Behavior (v0.4.3)

No MCP selection UI is available. The agent must infer the target server from conversation context, which is error-prone when tool names are identical across servers.

Environment

  • Qoder IDE version: 0.4.3
  • OS: Windows 22H2
  • MCP servers affected: apollo-*****1 and apollo-********2 (both using apollo-mcp-server with different endpoints)

Proposed Fix

Restore the MCP server selection UI that was available in earlier versions, allowing users to explicitly choose which server to route tool calls to when ambiguity exists.