A fork of @modelcontextprotocol/server-sequential-thinking that fixes the unreadable output problem in Claude Code CLI.
Claude Code v2.0.21+ prioritizes structuredContent over TextContent when rendering MCP tool output. The official sequential-thinking server returns both, causing the output to be displayed as a single-line raw JSON blob with escaped \n characters — making it nearly impossible to read.
See: claude-code#21186, claude-code#9962
This package simply removes the structuredContent return value, forcing Claude Code to fall back to the more readable TextContent rendering. No other changes — all thinking quality and features are preserved.
npm install -g @laststance/readable-sequential-thinkingclaude mcp add sequential-thinking -- npx @laststance/readable-sequential-thinkingOr in your MCP config:
{
"mcpServers": {
"sequential-thinking": {
"command": "npx",
"args": ["-y", "@laststance/readable-sequential-thinking"]
}
}
}Based on the original Sequential Thinking MCP Server by Model Context Protocol (LF Projects, LLC).
MIT