實際影片長度:3:27.000。原文、繁中、雙語可點擊句子跳轉影片。
0:00.000–0:09.560
what's going on everyone welcome back to this youtube channel in this video i'm going to show
0:09.560–0:12.233
you how you can connect codex with cloud code so
0:12.233–0:14.700
you can create sub agents that's going to be
0:14.700–0:17.163
spawned from your cloud code and gpt 5.
0:17.163–0:20.780
6 solve or whatever model you want from open ai is going to
0:20.780–0:23.659
do some tasks for you you can generate code with codex and
0:23.659–0:25.680
then review with cloud or generate code
0:25.680–0:28.584
with fable 5 and review it with gpt 5.6 solve so
0:28.584–0:31.640
it's an amazing combination that you can try and
0:31.640–0:36.100
the results are absolutely amazing in this video i'm going to show you how to connect them and if
0:36.100–0:40.024
you want me to make videos about how to use this connection afterwards then
0:40.024–0:41.100
do let me know down in
0:41.100–0:44.323
the comment section so without wasting any more time let's get started so
0:44.323–0:45.380
the first thing that you
0:45.380–0:49.920
need is the cloud code cli if you don't know how to install that i already have a video about that go
0:49.920–0:52.329
check that out the link will be in the description and
0:52.329–0:54.300
you need to have the codex i also have a video
0:54.300–0:55.601
about that you can go ahead and
0:55.601–0:58.360
check that after you have both of them what you will do is you will
0:58.360–1:05.140
run this command cloud mcp add codex dash dash codex mcp dash server just copy this command it will
1:05.140–1:07.335
also be down in the description box you can go ahead and
1:07.335–1:09.140
check it out from there and then hit enter and
1:09.140–1:14.520
you'll see that it says edit studio mcp server codex with command codex mcp dash server to local
1:14.520–1:19.700
config so so it just added this mcp server in cloud now we are going to use this mcp server to make
1:19.700–1:21.139
connection with codex and
1:21.139–1:24.280
we are going to get responses from the codex models so i can
1:24.300–1:28.915
also do cloud mcp list you can see it's checking the mcp server health and
1:28.915–1:30.300
it gives me all the mcp
1:30.300–1:35.520
servers that are connected with my cloud code you can see the codex is right here i also have the
1:35.520–1:38.216
magnific hicksfield gmail google notion but
1:38.216–1:41.480
this is the one that we are interested in now simply open
1:41.480–1:45.068
up the cloud let's ask it do you have access to codex and
1:45.068–1:47.620
you can see it says yes codex mcp server
1:47.620–1:52.441
connected two tools mcp codex codex mcp codex codex reply so
1:52.441–1:54.080
let's tell it can you
1:54.080–1:55.641
make codex say hello and
1:55.641–1:59.700
it asked me for a permission for the tool use i'm going to say yes
1:59.700–2:02.640
and you can see it says codex say hello so
2:02.640–2:06.560
we got the response from codex now i can tell it can you
2:06.560–2:11.486
run a subagent that writes a function to generate primes and
2:11.486–2:14.540
you review the code so i'm asking it to
2:14.540–2:17.050
create a subagent that will write a function so
2:17.050–2:19.560
gpt 5.6 is going to write that function and then
2:19.560–2:22.102
i'm asking it to be the code reviewer so
2:22.102–2:25.360
fable 5 is going to be our qa so if i hit enter you can
2:25.360–2:28.216
also see it says session thread alive with this tag and
2:28.216–2:30.500
can continue conversation with codex dash
2:30.500–2:32.850
triple i if want so it keeps the connection open so
2:32.850–2:35.200
we don't have to create a connection every time we
2:35.200–2:41.520
want to talk to codex after we start a connection now for our prime numbers program it's done now fable
2:41.520–2:46.220
five is reviewing the code you can see it's reading one file i'm using the caveman as well if you have
2:46.220–2:48.541
noticed going to give it the permission and
2:48.541–2:51.120
you can see it is now doing the qa so so it's saying
2:51.120–2:56.940
that subagent road scratchpad slash primes or py test pass review now it's reviewing the code and it's
2:56.940–2:58.401
telling us what is good and
2:58.401–3:02.120
what's not good which is really nice now you can use it to do anything
3:02.120–3:05.746
like you can generate some code with fable 5 and
3:05.746–3:09.280
then ask gpt 5.6 salt to review it and then keep
3:09.280–3:13.272
having that loop until uh the code is as per your needs until
3:13.272–3:15.960
you achieve your goals so yeah you can do
3:15.960–3:21.580
a lot of things with this uh let me know if you want to know what are the use cases and what we can
3:21.580–3:23.069
achieve with uh this connection so
3:23.069–3:25.740
i'll see you in the next one till then peace out and don't forget
3:25.740–3:27.000
to like and subscribe bye
0:00.000–0:09.560
大家好,歡迎回到這個 YouTube 頻道。在這支影片中,我將示範
0:09.560–0:12.233
如何將 Codex 與 Cloud Code 連接,以便
0:12.233–0:14.700
建立子代理程式,這些子代理程式將
0:14.700–0:17.163
由您的 Cloud Code 和 GPT 5.6 Solve 或其他您選擇的 OpenAI 模型
0:17.163–0:20.780
為您執行某些任務。您可以使用 Codex 產生程式碼,
0:20.780–0:23.659
然後使用 Cloud 進行審查,或使用 Fable 5 產生程式碼,
0:23.659–0:25.680
並使用 GPT 5.6 Solve 進行審查。因此
0:25.680–0:28.584
這是一個非常棒的組合,您可以嘗試看看,
0:28.584–0:31.640
結果絕對令人驚嘆。在這支影片中,我將示範如何將它們連接起來,如果
0:31.640–0:36.100
您希望我之後製作關於如何使用此連接的影片,請
0:36.100–0:40.024
在下方
0:40.024–0:41.100
留言區告訴我。那麼,不浪費更多時間,我們開始吧。所以
0:41.100–0:44.323
第一件事是您
0:44.323–0:45.380
需要 Cloud Code CLI。如果您不知道如何安裝,我已經有一支關於此的影片,請
0:45.380–0:49.920
查看,連結會在描述中,並且
0:49.920–0:52.329
您需要擁有 Codex。我也有一支關於此的影片,
0:52.329–0:54.300
您可以繼續查看。在您擁有兩者之後,您將
0:54.300–0:55.601
執行此指令:cloud mcp add codex --codex-mcp --server,只需複製此指令,它也會
0:55.601–0:58.360
在描述框中提供,您可以從那裡
0:58.360–1:05.140
查看,然後按下 Enter,您會看到它說編輯 Studio MCP 伺服器 Codex,指令為 codex-mcp --server 到本地
1:05.140–1:07.335
配置。所以它只是在 Cloud 中添加了這個 MCP 伺服器。現在我們將使用這個 MCP 伺服器來建立
1:07.335–1:09.140
與 Codex 的連接,並
1:09.140–1:14.520
從 Codex 模型獲取回應。所以我也可以執行 cloud mcp list,您可以看到它正在檢查 MCP 伺服器健康狀況,並
1:14.520–1:19.700
給我所有與我的 Cloud Code 連接的 MCP
1:19.700–1:21.139
伺服器。您可以看到 Codex 就在這裡,我還有
1:21.139–1:24.280
Magnific、Hicksfield、Gmail、Google、Notion,但
1:24.300–1:28.915
這是我們感興趣的。現在只需打開
1:28.915–1:30.300
Cloud,詢問它是否擁有 Codex 的存取權,
1:30.300–1:35.520
您可以看到它說「是」,Codex MCP 伺服器
1:35.520–1:38.216
已連接,兩個工具:mcp-codex、codex-mcp-codex、codex-reply。所以
1:38.216–1:41.480
讓我們告訴它,能否
1:41.480–1:45.068
讓 Codex 說「你好」,
1:45.068–1:47.620
它詢問我關於工具使用的權限,我將說「是」,
1:47.620–1:52.441
您可以看到它說「Codex 說你好」,所以
1:52.441–1:54.080
我們從 Codex 獲得了回應。現在我可以告訴它,能否
1:54.080–1:55.641
執行一個撰寫函數以產生質數的子代理程式,並
1:55.641–1:59.700
它要求我允許使用工具,我選擇同意
1:59.700–2:02.640
然後你可以看到它顯示 codex 說你好,所以
2:02.640–2:06.560
我們收到了來自 codex 的回應,現在我可以告訴它,你能
2:06.560–2:11.486
運行一個子代理來編寫生成質數的函數,並且
2:11.486–2:14.540
你來審查程式碼,所以我要求它
2:14.540–2:17.050
建立一個子代理來撰寫函式,所以
2:17.050–2:19.560
GPT 5.6 會撰寫該函式,然後
2:19.560–2:22.102
我要求它擔任程式碼審查者,所以
2:22.102–2:25.360
Fable 5 將擔任我們的 QA,如果我按下 Enter,你可以
2:25.360–2:28.216
看到它顯示此標籤的 session thread alive,並且
2:28.216–2:30.500
可以繼續與 codex-iii 對話,如果我想的話,所以它保持連線開啟,因此
2:30.500–2:32.850
我們不需要每次想要與 codex 對話時都建立連線,現在我們的質數程式已完成,現在 Fable
2:32.850–2:35.200
5 正在審查程式碼,你可以看到它正在讀取一個檔案,我也在使用 caveman,如果你已經
2:35.200–2:41.520
注意到,將授予它權限,
2:41.520–2:46.220
你可以看到它現在正在進行 QA,所以它說
2:46.220–2:48.541
子代理程式碼暫存區/primes 或 py test pass 審查,現在它正在審查程式碼,並且
2:48.541–2:51.120
告訴我們什麼是好的,
2:51.120–2:56.940
什麼是不好的,這真的很棒,現在你可以用它做任何事情,
2:56.940–2:58.401
例如,你可以用 Fable 5 產生一些程式碼,然後
2:58.401–3:02.120
要求 GPT 5.6 Salt 來審查它,然後繼續
3:02.120–3:05.746
保持這個循環,直到程式碼符合你的需求,直到
3:05.746–3:09.280
你達成目標,所以是的,你可以做
3:09.280–3:13.272
很多事情,嗯,如果你想知道使用案例以及我們可以
3:13.272–3:15.960
透過這個連線達成什麼,請告訴我,所以
3:15.960–3:21.580
我會在下一個影片見,到時再見,別忘了
3:21.580–3:23.069
按讚和訂閱,再見
3:23.069–3:25.740
我們下一集見,再見,祝好,別忘了
3:25.740–3:27.000
按讚和訂閱,再見
0:00.000–0:09.560
what's going on everyone welcome back to this youtube channel in this video i'm going to show
大家好,歡迎回到這個 YouTube 頻道。在這支影片中,我將示範
0:09.560–0:12.233
you how you can connect codex with cloud code so
如何將 Codex 與 Cloud Code 連接,以便
0:12.233–0:14.700
you can create sub agents that's going to be
建立子代理程式,這些子代理程式將
0:14.700–0:17.163
spawned from your cloud code and gpt 5.
由您的 Cloud Code 和 GPT 5.6 Solve 或其他您選擇的 OpenAI 模型
0:17.163–0:20.780
6 solve or whatever model you want from open ai is going to
為您執行某些任務。您可以使用 Codex 產生程式碼,
0:20.780–0:23.659
do some tasks for you you can generate code with codex and
然後使用 Cloud 進行審查,或使用 Fable 5 產生程式碼,
0:23.659–0:25.680
then review with cloud or generate code
並使用 GPT 5.6 Solve 進行審查。因此
0:25.680–0:28.584
with fable 5 and review it with gpt 5.6 solve so
這是一個非常棒的組合,您可以嘗試看看,
0:28.584–0:31.640
it's an amazing combination that you can try and
結果絕對令人驚嘆。在這支影片中,我將示範如何將它們連接起來,如果
0:31.640–0:36.100
the results are absolutely amazing in this video i'm going to show you how to connect them and if
您希望我之後製作關於如何使用此連接的影片,請
0:36.100–0:40.024
you want me to make videos about how to use this connection afterwards then
在下方
0:40.024–0:41.100
do let me know down in
留言區告訴我。那麼,不浪費更多時間,我們開始吧。所以
0:41.100–0:44.323
the comment section so without wasting any more time let's get started so
第一件事是您
0:44.323–0:45.380
the first thing that you
需要 Cloud Code CLI。如果您不知道如何安裝,我已經有一支關於此的影片,請
0:45.380–0:49.920
need is the cloud code cli if you don't know how to install that i already have a video about that go
查看,連結會在描述中,並且
0:49.920–0:52.329
check that out the link will be in the description and
您需要擁有 Codex。我也有一支關於此的影片,
0:52.329–0:54.300
you need to have the codex i also have a video
您可以繼續查看。在您擁有兩者之後,您將
0:54.300–0:55.601
about that you can go ahead and
執行此指令:cloud mcp add codex --codex-mcp --server,只需複製此指令,它也會
0:55.601–0:58.360
check that after you have both of them what you will do is you will
在描述框中提供,您可以從那裡
0:58.360–1:05.140
run this command cloud mcp add codex dash dash codex mcp dash server just copy this command it will
查看,然後按下 Enter,您會看到它說編輯 Studio MCP 伺服器 Codex,指令為 codex-mcp --server 到本地
1:05.140–1:07.335
also be down in the description box you can go ahead and
配置。所以它只是在 Cloud 中添加了這個 MCP 伺服器。現在我們將使用這個 MCP 伺服器來建立
1:07.335–1:09.140
check it out from there and then hit enter and
與 Codex 的連接,並
1:09.140–1:14.520
you'll see that it says edit studio mcp server codex with command codex mcp dash server to local
從 Codex 模型獲取回應。所以我也可以執行 cloud mcp list,您可以看到它正在檢查 MCP 伺服器健康狀況,並
1:14.520–1:19.700
config so so it just added this mcp server in cloud now we are going to use this mcp server to make
給我所有與我的 Cloud Code 連接的 MCP
1:19.700–1:21.139
connection with codex and
伺服器。您可以看到 Codex 就在這裡,我還有
1:21.139–1:24.280
we are going to get responses from the codex models so i can
Magnific、Hicksfield、Gmail、Google、Notion,但
1:24.300–1:28.915
also do cloud mcp list you can see it's checking the mcp server health and
這是我們感興趣的。現在只需打開
1:28.915–1:30.300
it gives me all the mcp
Cloud,詢問它是否擁有 Codex 的存取權,
1:30.300–1:35.520
servers that are connected with my cloud code you can see the codex is right here i also have the
您可以看到它說「是」,Codex MCP 伺服器
1:35.520–1:38.216
magnific hicksfield gmail google notion but
已連接,兩個工具:mcp-codex、codex-mcp-codex、codex-reply。所以
1:38.216–1:41.480
this is the one that we are interested in now simply open
讓我們告訴它,能否
1:41.480–1:45.068
up the cloud let's ask it do you have access to codex and
讓 Codex 說「你好」,
1:45.068–1:47.620
you can see it says yes codex mcp server
它詢問我關於工具使用的權限,我將說「是」,
1:47.620–1:52.441
connected two tools mcp codex codex mcp codex codex reply so
您可以看到它說「Codex 說你好」,所以
1:52.441–1:54.080
let's tell it can you
我們從 Codex 獲得了回應。現在我可以告訴它,能否
1:54.080–1:55.641
make codex say hello and
執行一個撰寫函數以產生質數的子代理程式,並
1:55.641–1:59.700
it asked me for a permission for the tool use i'm going to say yes
它要求我允許使用工具,我選擇同意
1:59.700–2:02.640
and you can see it says codex say hello so
然後你可以看到它顯示 codex 說你好,所以
2:02.640–2:06.560
we got the response from codex now i can tell it can you
我們收到了來自 codex 的回應,現在我可以告訴它,你能
2:06.560–2:11.486
run a subagent that writes a function to generate primes and
運行一個子代理來編寫生成質數的函數,並且
2:11.486–2:14.540
you review the code so i'm asking it to
你來審查程式碼,所以我要求它
2:14.540–2:17.050
create a subagent that will write a function so
建立一個子代理來撰寫函式,所以
2:17.050–2:19.560
gpt 5.6 is going to write that function and then
GPT 5.6 會撰寫該函式,然後
2:19.560–2:22.102
i'm asking it to be the code reviewer so
我要求它擔任程式碼審查者,所以
2:22.102–2:25.360
fable 5 is going to be our qa so if i hit enter you can
Fable 5 將擔任我們的 QA,如果我按下 Enter,你可以
2:25.360–2:28.216
also see it says session thread alive with this tag and
看到它顯示此標籤的 session thread alive,並且
2:28.216–2:30.500
can continue conversation with codex dash
可以繼續與 codex-iii 對話,如果我想的話,所以它保持連線開啟,因此
2:30.500–2:32.850
triple i if want so it keeps the connection open so
我們不需要每次想要與 codex 對話時都建立連線,現在我們的質數程式已完成,現在 Fable
2:32.850–2:35.200
we don't have to create a connection every time we
5 正在審查程式碼,你可以看到它正在讀取一個檔案,我也在使用 caveman,如果你已經
2:35.200–2:41.520
want to talk to codex after we start a connection now for our prime numbers program it's done now fable
注意到,將授予它權限,
2:41.520–2:46.220
five is reviewing the code you can see it's reading one file i'm using the caveman as well if you have
你可以看到它現在正在進行 QA,所以它說
2:46.220–2:48.541
noticed going to give it the permission and
子代理程式碼暫存區/primes 或 py test pass 審查,現在它正在審查程式碼,並且
2:48.541–2:51.120
you can see it is now doing the qa so so it's saying
告訴我們什麼是好的,
2:51.120–2:56.940
that subagent road scratchpad slash primes or py test pass review now it's reviewing the code and it's
什麼是不好的,這真的很棒,現在你可以用它做任何事情,
2:56.940–2:58.401
telling us what is good and
例如,你可以用 Fable 5 產生一些程式碼,然後
2:58.401–3:02.120
what's not good which is really nice now you can use it to do anything
要求 GPT 5.6 Salt 來審查它,然後繼續
3:02.120–3:05.746
like you can generate some code with fable 5 and
保持這個循環,直到程式碼符合你的需求,直到
3:05.746–3:09.280
then ask gpt 5.6 salt to review it and then keep
你達成目標,所以是的,你可以做
3:09.280–3:13.272
having that loop until uh the code is as per your needs until
很多事情,嗯,如果你想知道使用案例以及我們可以
3:13.272–3:15.960
you achieve your goals so yeah you can do
透過這個連線達成什麼,請告訴我,所以
3:15.960–3:21.580
a lot of things with this uh let me know if you want to know what are the use cases and what we can
我會在下一個影片見,到時再見,別忘了
3:21.580–3:23.069
achieve with uh this connection so
按讚和訂閱,再見
3:23.069–3:25.740
i'll see you in the next one till then peace out and don't forget
我們下一集見,再見,祝好,別忘了
3:25.740–3:27.000
to like and subscribe bye
按讚和訂閱,再見

影片筆記:Connect Codex With Claude Code | Run fable 5 and GPT-5.6 Sol together

一句話總結

本影片演示如何透過 MCP (Model Context Protocol) 伺服器將 Codex 串接至 Cloud Code,並建立由 Cloud Code 生成的子代理 (Sub-agents),利用 GPT 5.6 進行程式碼生成,並由 Fable 5 擔任 QA 進行程式碼審查的循環流程。

核心重點

  1. 串接架構:透過 cloud mcp add 指令將 Codex 加入 Cloud Code 的本地設定,建立 MCP 伺服器連線。
  2. 子代理機制:在 Cloud Code 介面中,可建立子代理 (Sub-agents) 來執行特定任務。
  3. 生成與審查循環
  • 使用 GPT 5.6 (或稱 gpt 5.6 solve) 負責生成程式碼(例如:生成質數函數)。
  • 使用 Fable 5 負責擔任 QA/程式碼審查者。
  • 審查過程包含讀取檔案、執行測試(如 primes or py test pass review)並提供優缺點評估。
  1. 連線持續性:系統顯示 session thread alive 標籤,表示連線持續開啟,避免每次對話都需重新建立連線。
  2. 驗證方式:透過 cloud mcp list 檢查伺服器健康狀態,並執行 codex say hello 測試工具存取權。

詳細大綱

1. 前置準備

  • 安裝 Cloud Code CLI
  • 安裝 Codex

2. 建立 MCP 伺服器連線

  • 執行指令將 Codex 加入 Cloud Code 的本地設定:
  • cloud mcp add codex --codex-mcp-server
  • 檢查連線狀態與伺服器清單:
  • 執行 cloud mcp list 查看已連線的伺服器及健康狀態。
  • 確認工具列表中是否包含 mcp-codex 等項目。

3. 驗證連線與工具使用

  • 在 Cloud Code 介面確認具備 Codex 存取權。
  • 執行測試指令 codex say hello
  • 授予工具使用權限以獲取回應。

4. 實作子代理與程式碼審查循環

  • 任務設定:要求建立子代理撰寫生成質數 (primes) 的函數。
  • 模型分工
  • GPT 5.6:負責生成程式碼。
  • Fable 5:負責擔任 QA/程式碼審查。
  • 審查流程
  • 系統顯示 session thread alive,確保連線持續。
  • 審查者讀取檔案。
  • 執行測試指令(如 primes or py test pass review)。
  • 提供優缺點評估。
  • 應用:透過此循環機制持續生成與審查程式碼,直到符合需求。

5. 結語

  • 邀請觀眾留言詢問具體使用案例。

工具 / 模型 / 名詞整理

  • Cloud Code:開發環境/工具。
  • Codex:AI 程式碼生成工具。
  • Cloud Code CLI:Cloud Code 的命令列介面。
  • MCP (Model Context Protocol) Server:模型情境協議伺服器,用於串接不同工具。
  • GPT 5.6 (或 gpt 5.6 solve):影片中指定的用於程式碼生成的模型。
  • Fable 5 (或 fable 5):影片中指定的用於 QA/程式碼審查的模型。
  • OpenAI:GPT 模型相關公司。
  • Edit Studio MCP Server:指令輸出中提及的伺服器名稱。
  • Magnific Hicksfieldcloud mcp list 輸出中出現的伺服器名稱。
  • Gmail, Google, Notion, YouTubecloud mcp list 輸出中出現的其他伺服器名稱。

操作流程整理

  1. 安裝階段
  • 安裝 Cloud Code CLI。
  • 安裝 Codex。
  1. 連線設定階段
  • 執行 cloud mcp add codex --codex-mcp-server
  • 執行 cloud mcp list 確認連線成功。
  1. 功能測試階段
  • 在介面確認 mcp-codex 等工具已連線。
  • 執行 codex say hello 並授予權限測試回應。
  1. 實作應用階段
  • 建立子代理。
  • 輸入指令要求生成質數函數。
  • 系統自動分配 GPT 5.6 生成程式碼。
  • 系統自動分配 Fable 5 進行審查(讀檔、測試、評估)。
  • 確認 session thread alive 狀態,維持連線進行後續迭代。

值得注意的限制或風險

  • 模型名稱非標準:影片中使用的 GPT 5.6gpt 5.6 solveFable 5 並非目前公開標準的模型版本名稱,可能為內部版本、聽寫錯誤或特定實驗性模型。
  • 指令辨識錯誤風險:部分指令片段(如 codex dash triple icodex dash triple i)疑似聽寫錯誤,實際操作時需查證正確參數。
  • 路徑/指令語意不明:審查結果中出現 subagent road scratchpad slash primes or py test pass review,語意不清,可能影響實際操作復現。

逐字稿辨識疑點

  • gpt 5.6 solve:逐字稿中多次出現此名稱,疑似為模型版本或特定功能的稱呼,但非標準公開模型名稱,需查證。
  • fable 5:逐字稿中作為與 GPT 5.6 搭配使用的模型名稱,需查證是否為特定內部模型或聽寫錯誤。
  • codex mcp dash server:指令 cloud mcp add codex dash --codex mcp dash server 中的 dash 疑似為破折號 - 的聽寫錯誤,但依規則保留原樣。
  • codex dash triple i:指令片段 codex dash triple i 疑似為參數或標記的聽寫錯誤,需查證。
  • subagent road scratchpad slash primes or py test pass review:審查結果中的路徑或指令 road scratchpad slash primes or py test pass review 語意不明,疑似聽寫錯誤,需查證。
  • magnific hicksfield gmail google notion:在 cloud mcp list 輸出的伺服器清單中,magnific hicksfieldgmailgooglenotion 並列,magnific hicksfield 疑似為特定 MCP 伺服器名稱或聽寫錯誤,需查證。
  • edit studio mcp server:指令執行後的輸出訊息 edit studio mcp server,需查證是否為正確的產品或服務名稱。

可延伸追問

  • Fable 5GPT 5.6 的具體技術規格或來源為何?
  • 如何正確拼寫 codex dash triple icodex dash triple i 等指令參數?
  • Magnific Hicksfield 是什麼樣的 MCP 伺服器?
  • 如何自訂 Edit Studio MCP Server 或其他列出的伺服器?
  • 在實際開發中,如何確保 session thread alive 的穩定性以避免連線中斷?

尚未產生學習筆記

請在 Telegram 指令最後加上「學習」,例如:videonote 網址 英文 雙語 學習