reddit · SideProject · 10-13 21:50 51 分 ← 返回列表

Mirador: a canvas for coding agents, so it draws what it means instead of writing walls of text

I use Claude Code every day, and what wore me down was the reading. Every plan, explanation and review arrives as a wall of text in a terminal, and I'd catch myself skimming the one part that mattered. So I built Mirador. It runs the Claude Code (or opencode) you already have, with your own sign-in, settings, hooks and MCP servers, and puts a canvas next to the chat. The agent gets a few tools for it: show : draws what it means as a page, such as a sequence diagram, a moc

五维评分

付费意愿

需求真实度

个人可落地

竞争空间

新鲜度

AI 分析

痛点：每天使用终端 AI 编程代理的开发者被迫阅读大段文字输出的计划、解释和 review，需要可视化/结构化呈现来降低理解成本。

目标用户：每天使用 Claude Code、opencode 等终端 AI 编程代理的开发者与团队。

证据：what wore me down was the reading. Every plan, explanation and review arrives as a wall of text in a terminal, and I'd catch myself skimming the one part that mattered.

变现：桌面客户端订阅制（按席位/月），团队协作版（review、PR 评论同步）加价，或与 Claude Code/opencode 生态做集成付费插件。

MVP：不做完整桌面端，先做一个把 agent 输出渲染成序列图/线框图/checklist 的轻量 MCP server + 浏览器单页画布，接入 Claude Code，一周验证用户是否愿意用它替代终端阅读。

竞品：Cursor、Cline、Warp、Mermaid CLI/终端图表工具、Claude Code 自带 diff/plan UI

摘要：Mirador 是作者为 Claude Code/opencode 做的可视化画布，用 show/ask/tour/review/checklist 工具把终端里的大段文字变成图表、选择预览和代码导览，解决 AI 编程代理输出难读的痛点。已在 macOS/Linux 发布，通过 ACP + MCP + WebSocket 实现，源码开源。作者只表达个人痛点，无收入或付费意愿数据。