Back to Details

supacortex

Compare original and translation side by side

🇺🇸

Original

English
🇨🇳

Translation

Chinese

Supacortex — CLI Skill

Supacortex — CLI 技能

Supacortex is a personal memory layer. Two CLI commands: bookmarks and conversations.
Supacortex是一个个人记忆层，包含两个CLI命令：书签（bookmarks）和对话（conversations）。

Setup

安装配置

bash
npm i -g @supacortex/cli
scx login
bash
npm i -g @supacortex/cli
scx login

1. Bookmarks (
scx bookmarks
)

1. 书签（
scx bookmarks
）

Save and search bookmarks (links, YouTube videos).
保存和搜索书签（链接、YouTube视频）。

Commands

命令

List bookmarks

列出书签

bash
scx bookmarks list [--search "<query>"] [--type <tweet|link|youtube>] [--limit <n>] [--offset <n>] [--pretty]
bash
scx bookmarks list [--search "<query>"] [--type <tweet|link|youtube>] [--limit <n>] [--offset <n>] [--pretty]

Add a bookmark

添加书签

bash
scx bookmarks add <url> [--pretty]
bash
scx bookmarks add <url> [--pretty]

Get a bookmark by ID

根据ID获取书签

bash
scx bookmarks get <id> [--pretty]
bash
scx bookmarks get <id> [--pretty]

Delete a bookmark

删除书签

bash
scx bookmarks delete <id> [--pretty]
bash
scx bookmarks delete <id> [--pretty]

When to use bookmarks

书签使用场景

  • User asks to save a link or YouTube video
  • User wants to search their saved content
  • When you need to reference previously saved URLs
  • 用户要求保存某个链接或YouTube视频
  • 用户想要搜索自己已保存的内容
  • 你需要引用之前保存过的URL时

2. Conversations (
scx conversation
)

2. 对话（
scx conversation
）

Save summaries of AI chat sessions. Every conversation has a tier that determines its depth.
保存AI聊天会话的摘要，每段对话都有一个**层级（tier）**来决定其记录深度。

Tiers

层级说明

TierWhen to useContent format
brief
Throwaway queries, quick lookupsSingle sentence: "Asked about JSON parsing in Bun"
summary
Most working sessionsMarkdown — 3-8 bullet points covering what was discussed, decided, and found
detailed
Deep sessions with architectural decisions, research findingsMarkdown — full structured document with headings, reasoning, code snippets, follow-ups
层级使用场景内容格式
brief
一次性查询、快速查找单句："Asked about JSON parsing in Bun"
summary
大多数工作会话Markdown — 3-8个要点，覆盖讨论内容、决策结果和查询到的信息
detailed
涉及架构决策、研究结论的深度会话Markdown — 带有标题、推理过程、代码片段、后续待办的完整结构化文档

Commands

命令

Save a conversation

保存对话

bash
scx conversation add "<content>" --tier <brief|summary|detailed> [--title "<title>"] [--metadata '<json>'] [--pretty]
The 
--tier
flag is required. It maps to memory types: 
conversation_brief
, 
conversation_summary
, 
conversation_detailed
.
Examples:
bash
undefined
bash
scx conversation add "<content>" --tier <brief|summary|detailed> [--title "<title>"] [--metadata '<json>'] [--pretty]
--tier
参数为必填项，它对应记忆类型：
conversation_brief
、
conversation_summary
、
conversation_detailed
。
示例：
bash
undefined

Brief — one sentence

简短模式 — 单句

scx conversation add "Helped debug CORS issue in Hono API" --tier brief
scx conversation add "Helped debug CORS issue in Hono API" --tier brief

Summary — bullet points

摘要模式 — 要点列表

scx conversation add "- Set up memory table with tsvector search
  • Added triggers for auto search vector generation
  • Created API routes for CRUD
  • Decided on hybrid schema approach"
    --tier summary
    --title "Memory system setup"
    --metadata '{"source": "claude-code"}'
scx conversation add "- Set up memory table with tsvector search
  • Added triggers for auto search vector generation
  • Created API routes for CRUD
  • Decided on hybrid schema approach"
    --tier summary
    --title "Memory system setup"
    --metadata '{"source": "claude-code"}'

Detailed — full document

详细模式 — 完整文档

scx conversation add "## Memory Architecture Decision..." --tier detailed --title "Memory layer brainstorm"
undefined
scx conversation add "## Memory Architecture Decision..." --tier detailed --title "Memory layer brainstorm"
undefined

List conversations

列出对话

bash
scx conversation list [--search "<query>"] [--tier <brief|summary|detailed>] [--limit <n>] [--offset <n>] [--pretty]
bash
scx conversation list [--search "<query>"] [--tier <brief|summary|detailed>] [--limit <n>] [--offset <n>] [--pretty]

Get a conversation by ID

根据ID获取对话

bash
scx conversation get <id> [--pretty]
bash
scx conversation get <id> [--pretty]

Update a conversation

更新对话

bash
scx conversation update <id> [--title "<title>"] [--content "<content>"] [--tier <tier>] [--metadata '<json>'] [--pretty]
bash
scx conversation update <id> [--title "<title>"] [--content "<content>"] [--tier <tier>] [--metadata '<json>'] [--pretty]

Delete a conversation

删除对话

bash
scx conversation delete <id> [--pretty]
bash
scx conversation delete <id> [--pretty]

When to save conversations

对话保存场景

Save when the user says:
  • "save to cortex" / "save to supacortex"
  • "save this session" / "remember this"
  • "log this conversation"
当用户说出以下内容时保存：
  • "save to cortex" / "save to supacortex"
  • "save this session" / "remember this"
  • "log this conversation"

When to recall conversations

对话调取场景

Pull past conversations when the user says:
  • "check my conversation about X"
  • "pull the X conversation"
  • "get the summary for X"
  • "what did we work on last time?"
当用户说出以下内容时拉取过往对话：
  • "check my conversation about X"
  • "pull the X conversation"
  • "get the summary for X"
  • "what did we work on last time?"

Metadata

元数据

Metadata is freeform JSON passed via 
--metadata
. The AI decides what to store. Common fields:
  • source
    — where this was captured ("claude-code", "chatgpt", "opencode")
  • tags
    — array of topic tags
  • project
    — which project the conversation was about (e.g. "supacortex", "supalytics")
  • category
    — topic area: "project", "life", "general", "learning", "work"
元数据是通过
--metadata
参数传入的自由格式JSON，由AI决定存储的内容，常见字段如下：
  • source
    — 内容捕获来源（"claude-code"、"chatgpt"、"opencode"）
  • tags
    — 主题标签数组
  • project
    — 对话所属的项目（例如"supacortex"、"supalytics"）
  • category
    — 主题领域："project"、"life"、"general"、"learning"、"work"

Consistent verbs across all commands

所有命令统一操作动词

Both commands (
bookmarks
, 
conversation
) support: 
list
, 
add
, 
get
, 
delete
.
conversation
also supports: 
update
.
All commands output JSON by default (optimized for AI agents). Use 
--pretty
for human-readable output.
两个命令（
bookmarks
、
conversation
）都支持：
list
、
add
、
get
、
delete
。
conversation
命令还额外支持：
update
。
所有命令默认输出JSON（为AI Agent优化），使用
--pretty
参数可以输出人类易读的格式。