Open a Web UI to collect interactive feedback from the user.

This tool is the PRIMARY communication channel with the user. Route substantive content (analyses, plans, code summaries, questions needing confirmation) through this tool's summary parameter rather than the Cursor chat, because the Web UI renders Markdown far better.

MANDATORY call scenarios — do NOT end the turn without calling: 1. After any code modification — summarize the change, wait for the user to test. 2. Before implementing a new requirement — explain your understanding and approach, wait for confirmation. 3. When a tool call is cancelled/interrupted — ask the user the next step. 4. When this MCP call times out — immediately re-call; do not assume the user has left. 5. When the user explicitly skips without giving feedback — immediately re-call and ask why they skipped. 6. At the end of any multi-step task stage — check in. 7. Whenever you would otherwise ask the user a question in plain chat — route it through this tool's summary instead.

STRICT PROHIBITIONS: - While this tool is available, DO NOT invoke Cursor's built-in AskQuestion / question-picker tool. Every question to the user MUST go through this tool's summary (ask with a numbered list inside the Markdown). - DO NOT end your turn with a plain-text reply when you still have output intended for the user to read. Put that output into summary and call this tool instead. - Allowed parameters: project_directory, summary, title, timeout. title is OPTIONAL — pass a short headline (≤30 chars) for the Web UI's tab title and summary header. Names such as question, choices, options belong to AskQuestion — NEVER forward them to this tool. Passing any other keyword argument will cause the call to fail schema validation on the server.

Only stop calling when the user explicitly says 结束 / 不用了 / stop / done or an equivalent termination phrase.

summary writing rules: - Write in the same language the user is currently using. - Use Markdown: headings (##), lists, tables, code fences for snippets, bold for key decisions. - Include modified file paths, key trade-offs, and explicit questions for the user when applicable. - NEVER pass an empty string or the placeholder phrase 我已完成了您请求的任务。. The tool rejects such values at runtime and returns a correction notice that you must follow by re-calling with a real summary.

title writing rules: - Optional. Leave empty when summary is self-explanatory. - Keep concise (≤30 chars) — it's a headline, not a sentence. - Use the same language as the user. - Examples: 需要你提供更多信息 / 修复完成，请验证 / 实现方案确认 / Need more context.

Args: project_directory: Absolute or relative project path. Defaults to cwd. summary: Required. Markdown summary of the AI's current work / plan / question. No empty string, no placeholder phrase. title: Optional short headline shown in the Web UI tab and summary header. Leave empty if there's no concise headline. timeout: Seconds to wait for user feedback. Keep >= 600.

Returns: ToolResult: TextContent with the user's text feedback plus MCPImage objects for any uploaded images. An empty result means the user skipped — you MUST re-call this tool to ask why.

获取系统环境信息

Returns: str: JSON 格式的系统信息