Turning "Smart-Only AI" into a "Company Colleague". Stories of 8 MCPs Used Daily

AI translation

Supabase, Playwright, Tavily... I'll honestly introduce the MCPs I use every day in development and operations with real examples.

About 2 min read63

Show original

Enjoyed this article?

Use "Request tipping" to ask the author to set up tip receiving.

Share

平山耕司Published June 19, 2026

As I've been having AI help me with code and research every day, I started wondering: what's the difference between "an AI that's just smart" and "an AI that actually takes action"?

What bridged that gap for me was MCP. It's a mechanism that gives intelligent AI "hands and feet"—enabling it to touch Slack, production databases, and browsers.

I wrote honestly about 8 MCPs I actually use daily in development and operations on note. For example:

AI directly reads the production database to investigate the root cause of bugs (Supabase / with a "read-only" boundary)
Completely takes over browser operations (Playwright / downloading invoices, etc.)
Returns search results in a format that's easy for AI to use (Tavily)
And integrates Slack, Notion, GitHub, meeting notes, etc. to pass "company context" to AI—a custom MCP I built

That last custom MCP is TeamLog, which we're building. (...omitted...) An AI that's just smart gradually transforms into a company "colleague." I'd be happy if this becomes a reference or decision-making material for those who are similarly making the most of AI in personal projects or small teams.

▼ Full text
→ note: https://note.com/hirayama_takea/n/n2ea08465b017

▼ TeamLog (also listed on tsukutta / Free plan available)
→ https://manemate.takea.co.jp/?utm_source=tsukutta&utm_medium=referral&utm_campaign=tsukutta-mcp-summary-2026-06&utm_content=article-cta

AIに毎日コードや調べ物を手伝ってもらっているうちに、「賢いだけのAI」と「実際に手を動かしてくれるAI」の差はどこにあるんだろう、と考えるようになりました。

その差を埋めてくれたのが MCP でした。賢いだけのAIに「手足」を生やして、Slackも本番のデータベースもブラウザも触れるようにする仕組みです。

noteに、僕が毎日の開発・運用で実際に使い込んでいるMCPを8個、正直に書きました。たとえば、

本番のデータベースをAIが直接読んで、不具合の原因を調べる（Supabase / ただし「読むだけ」の境界つき）
ブラウザ操作をまるごと肩代わり（Playwright / 請求書のダウンロードなど）
検索結果を、AIが使いやすい要点の形で返す（Tavily）
そして、Slack・Notion・GitHub・議事録などを統合して「会社の文脈」をAIに渡す、自作のMCP

最後の自作MCPが、僕らが作っている TeamLog です。（…中略…）賢いだけのAIが、だんだん会社の"同僚"に変わっていく。同じように個人開発や小さなチームでAIを使い倒している方の参考や、判断材料になればうれしいです。

▼ 全文
→ note: https://note.com/hirayama_takea/n/n2ea08465b017

▼ TeamLog（tsukuttaにも掲載 / Freeプランあり）
→ https://manemate.takea.co.jp/?utm_source=tsukutta&utm_medium=referral&utm_campaign=tsukutta-mcp-summary-2026-06&utm_content=article-cta