Non-ASCII characters returned as \uXXXX escape sequences instead of UTF-8

[iceliu]

When reading messages from Slack channels via the MCP server (https://mcp.slack.com/mcp), all non-ASCII characters (Chinese/Japanese/Korean, etc.) are returned as JSON Unicode escape sequences (\uXXXX) instead of decoded UTF-8 text.

Example

The slack_read_channel tool returns:

{"messages": "Channel: #\u5c08\u6848\u8a0e\u8ad6\n\n=== Message from \u674e\u5927\u660e ===\n\u660e\u5929\u4e0b\u5348\u4e09\u9ede\u958b\u6703\uff0c\u8acb\u5927\u5bb6\u6e96\u6642\u53c3\u52a0"}

Expected (UTF-8):

{"messages": "Channel: #專案討論\n\n=== Message from 李大明 ===\n明天下午三點開會，請大家準時參加"}

Impact

AI agents consuming this output must infer CJK characters from hex codepoints, which leads to frequent misreadings of names and content — especially problematic for languages like Chinese where a single wrong codepoint produces a completely different character.

Suggested fix

Serialize JSON responses with ensure_ascii=false (or equivalent) so non-ASCII characters are output as UTF-8 directly.

[stevengill]

Thanks for send the bug report. I will pass it on to the team.