Free to read · no account needed

10 min · plugins

Build vs install a plugin

Goal: Write the job story and “must never happen,” then compare three install options—do not start by building MCP.

Best for: when someone says “we should build an MCP server” on week one.

Most beginners should install/configure first. Build only when a needed action does not exist, or compliance blocks third-party tools. This page is a decision filter—not a tutorial on writing servers. When you do need a first safe server later, see Safe first MCP server. For evaluating installs, see Install and evaluate.

Scene tip: Decide on a Tuesday when you feel the urge to “Build vs install a plugin” Buy/eval first.

Why beginners skip this

Building when a ready plugin exists feels like control—maintenance eats this week’s progress. Ask “who maintains?” first.

Step 1 — Write the job story

Write one paragraph: who, goal, which apps touch, how often, what must never happen—wish lists are not requirements.

Turn this wish into a job story (≤8 lines):
Wish: [paste]
Include a never-happen list (delete, wrong folder, external send…).
Example

Filled prompt

Wish: auto-file invoices from email into Drive folders by vendor.
Job story + never-happen (delete mail, wrong vendor folder, send replies).

What good output looks like

  • Eval score written before install for “Build vs install a plugin”
  • Uninstall path and scopes listed
  • No always-allow on write/send
  • Pass/fail: keep-or-remove decision dated

Step 2 — Search existing options

Search built-in connectors and official plugins before random stores; score fit 1–5—official first.

Job story:
[paste]
Propose 3 “buy/install” options to investigate (categories OK if names unknown).
For each: what to verify in 10 minutes (permissions, last update, read-only test).
Example

Filled prompt

Job: auto-label support mail by product line (3 labels). Weekly volume ~40 threads.
Write build-vs-buy card: frequency, security (no auto-send), eval task (label 5 samples read-only), install candidate, build only if….
Decide in 6 bullets. No romance about "our own plugin."

What good output looks like

  • Uninstall path and scopes listed
  • No always-allow on write/send
  • Pass/fail: keep-or-remove decision dated
  • Eval score written before install for “Build vs install a plugin”

Step 3 — Build only with a trigger

Build only when a needed action is missing, data cannot leave the boundary, or install options failed twice—one action is enough.

Should we build? Answer yes/no with 5 bullets using this evidence: [paste eval notes].
If yes, define the smallest build (one action). If no, say which installs to combine.
Example

Filled prompt

Need: weekly release-note draft from merged PR titles.
Eval two installs on 5 sample PRs before any build. Human publishes.
6-bullet buy/build card.

What good output looks like

  • No always-allow on write/send
  • Eval score written before install for “Build vs install a plugin”
  • Pass/fail: keep-or-remove decision dated
  • Uninstall path and scopes listed

Keep it real

Run this guide’s Practice now on work you will actually ship this week—not a fictional scenario. Dated notes beat perfect theory.

Do not bookmark-only

Finish Practice now, then tick this guide on your path. Titles unread are not skills.

Close the loop

Write one line: next time I build with “who maintains?” blank, I refuse to start.

Red flag

Building before “must never happen”; vibes over three install options—stop the scaffold.

Save this card

Card: Write “must never happen,” then compare three install options—do not build on vibes. Keep this beside the tool; read before you start.

Before you start

Open the listing and a notes doc. Score purpose/scopes/uninstall for “decide whether an existing connector is enough” before Install. Prefer wait over Always allow. Guide: plugins-when-to-build-vs-buy. Keep the real job visible the whole time; a blank box means stop and fetch the artifact first.

How to use this week

This week write a three-line build-vs-buy compare for a real need; do not build if “who maintains?” is blank.

Close the week by quoting aloud the refuse line you actually used once. If you cannot, rerun Practice now and date the pass/fail (Asia/Shanghai).

Pitfalls

PitfallFix
Building for resume pointsInstall first
Custom tool with wide secretsLeast privilege
No job storyWrite it before shopping
“MCP” as fashion, not needRequire a missing-action trigger
Example

Filled prompt

Temptation: skip the habit for plugins-when-to-build-vs-buy to save five minutes.
Job still: decide whether an existing connector is
Smallest correct pass: Done visible · checkable artifact · Unknowns [?].
Boundaries: draft only; no silent send; no invented metrics.
Return: artifact + refuse-line said aloud.

What good output looks like

  • Names the shortcut refused (specific to plugins-when-to-build-vs-buy)
  • Still delivers checkable output for “decide whether an existing connector is”
  • Unknowns labeled; draft-only
  • Pass/fail: refuse-line dated (plugins-when-to-build-vs-buy #4)

Practice now

TCO line: 90-day cost estimate on Save (Asia/Shanghai).

Sunset clause: if build, write who deletes it in 90 days (Asia/Shanghai).

Build-vs-buy drill: who maintains? cost? blast radius?—blank “who maintains” means buy/wait (Asia/Shanghai).

Thin-spot drill: rewrite the Save card with real nouns from your work (repo, doc, inbox). Generic placeholders fail the drill.

Tonight write a three-line build-vs-buy compare for a real need; do not build if “who maintains?” is blank.

Today card (copy) — plugins-when-to-build-vs-buy: 1. Artifact I will produce: ___ 2. Pass/fail check: ___ 3. Refuse-line if tempted to skip verify/send: ___ (date: Asia/Shanghai) 4. One never-do from this guide: ___

Extra pass: do the card on a real file today—reading alone does not count.

Ship rule: name the buyer vs build criterion you used in one sentence; if you cannot, do not install yet.

Buffer line: name the artifact path you used (Asia/Shanghai).

Ship-check: artifact path ___ · refuse-line dated Asia/Shanghai ___ · never-do ___

One sentence takeaway

Build-vs-buy three-line compare; decide only after “who maintains?” is clear.

Next

Safe first MCP server · Install and evaluate · Directory hygiene

← Back to home · Learn · Tools · Compare · Start