🚀 From our VS Code Learn library: Extending Agents! In this E2E tutorial, we'll explore key aspects of AI Agents with Tools, Tool sets, Permissions and Sandboxing. We'll also look at MCP (Model Context Protocol) and how to use it to connect your agents to external data sources. Check it out ▶️ piped.video/SN3JTq3Q9-M

Sep 24, 2026 · 10:05 PM UTC

14
16
151
26,850
Sort replies: Relevant Recent Liked
Replying to @code
ファーストビューが命だけど、実はUSPが脇役じゃない？ 「サンドボクシング」 →権限や環境を限定し安全に動作させる仕組み 脳は情報の安全地帯を感じると理解が深まるから 強力なツールの裏には確かな信頼感があるということ LPは料理の盛り付けと味の関係みたいなものですね
1
27
Replying to @code
The permissions part is the one people skip. Once one agent hands work to a sub-agent, does the sub-agent inherit the full tool set, or can you narrow it per call?
15
Replying to @code
League of Legends
@LeagueOfLegends
10 Nov 2019
WAKE UP
66
Replying to @code
the sandboxing bit is the part people skip until an agent rm -rf's something stupid. permissions first, tools second.
122
Replying to @code
比起接多少工具，我更想看权限和沙箱怎么配。
52