20 Global Variables to Set Up Today

TL;DR: AI prompt variables are values you save once — a company name, a tone, a support email — and reuse across every prompt with {{variable_name}} syntax, instead of retyping the same details every session. Prompt Architects calls this feature Global Variables. Below are 20 worth setting up today, grouped by what they're for, plus the one rule to follow before you save your first one: never put anything in a variable you wouldn't want sent to an AI model on every request.

What is an AI prompt variable, exactly?

A global variable is a value you define once in your Prompt Architects account and reuse across prompts by writing its name inside double curly braces, like {{company_name}} or {{tone}}. You give it a short id, a name, and a value on the Variables page. From that point on, any prompt containing {{company_name}} gets replaced with whatever you saved, at the moment you run the prompt, not before, and not baked permanently into the prompt text.

That distinction matters more than it sounds. The prompt itself stays generic. If you write "Write a tweet for {{company_name}} in a {{tone}} tone," the saved prompt keeps working after you rebrand, change jobs, or start writing for a different client. You just update the variable once, and every prompt that references it picks up the new value on its next run. Nothing about the prompt text itself needs to change.

This is a genuinely different thing from typing [company name] as a placeholder and manually swapping it in yourself. A manual placeholder is a habit you maintain. A global variable is account state: it persists, it syncs, and it applies itself. The gap between the two is exactly the gap between remembering to do something and not having to remember at all.

Variables sync to your account, so they follow you across the web app and the Chrome extension, and across every AI platform the extension supports, including ChatGPT, Claude, Gemini, Grok, and several others. They also work through the Prompt Architects MCP server inside Claude Desktop, Cursor, Claude Code, and Codex, so the same values are available from an IDE, not only from a browser tab. If you're already keeping a personal prompt library, variables are the layer that stops that library from going stale every time a detail about your business changes.

Most people set up an AI tool once, on day one, and never open its settings again. Global Variables is the kind of feature that sits quietly in that gap: genuinely useful, rarely visited, and worth five minutes today precisely because there's no natural moment later that will remind you to come back to it.

What should never go in a variable?

Everything below is written with that filter already applied: names, roles, tone words, links. None of it is the kind of information this warning is about, and that's deliberate. Think of the list less as "20 things to type in" and more as 20 categories worth checking your own information against before you save anything.

How do you create a global variable?

Setting one up takes under a minute:

1. Open the web app or the Chrome extension and go to the Variables section.
2. Click New variable.
3. Give it a short id (letters, numbers, and underscores, since this is what goes inside the {{ }}), a display name, and a value.
4. Save.

The variable is available immediately, in every prompt you write, on every device where you're signed in. When you're typing a prompt and type {{, an autocomplete panel shows your existing variables so you don't need to remember every id you've created. That gets genuinely useful once you're past three or four of them, since typing the first couple of characters is usually enough to find the one you want.

Editing an existing variable works the same way in reverse: open it, change the value, save, and every prompt that already references it starts using the new value the next time you run it. This is the whole point of setting one up in the first place. If you rebrand, change jobs, switch which product you write about most, or just realize your {{tone}} value has been slightly wrong for a month, you fix it once instead of hunting down every prompt that used the old version.

20 global variables worth setting up today

None of these need to be exotic. The unlock isn't a clever variable; it's having the boring, stable facts about your work saved once instead of retyped in every session. Here are 20, grouped by what they cover.

Brand and identity (4)

{{company_name}} = Acme Corp
{{tagline}} = Project management that doesn't need a manual
{{elevator_pitch}} = A 30-second description of what Acme Corp does and for whom
{{brand_voice}} = Direct, a little dry, no exclamation points

These four show up in almost every piece of external writing: emails, ad copy, landing pages, social posts. Instead of re-describing your company every time, reference {{company_name}} and {{brand_voice}} and let the model carry the description forward consistently. The elevator pitch is worth writing carefully once, since a rushed one-liner written under deadline pressure tends to get reused unedited for months.

Audience and positioning (3)

{{target_audience}} = Ops managers at 20-100 person logistics companies
{{icp}} = Series A-B SaaS founders who've hired their first ops hire
{{competitor_name}} = Whatever tool you most often get compared to

Useful for anything comparative or persuasive: landing page copy, cold outreach, competitive positioning. Update {{icp}} once when your ideal customer shifts, and every prompt that references it stops targeting last quarter's buyer. {{competitor_name}} is the one to keep current if you're in a fast-moving category; a stale comparison target makes an otherwise sharp piece of copy look dated the moment someone recognizes the name has changed.

Tone and style (3)

{{tone}} = Warm but efficient, like a colleague who respects your time
{{reading_level}} = Plain language, no jargon, aim for a 9th-grade reading level
{{style_reference}} = Short sentences, active voice, one idea per paragraph

These three are what make a rewritten draft actually sound like you instead of sounding like "AI wrote this." Store the tone you actually want, not the tone you assume the model will default to, and be specific: "warm but efficient" gives the model something to work with, where "professional" alone tends to produce the same flat, cautious register every tool defaults to on its own.

Role and work context (3)

{{job_title}} = Head of Customer Success
{{department}} = Customer Success
{{manager_name}} = The person whose updates you write for, if that changes rarely

If you write status updates, performance reviews, or internal memos regularly, these save you from re-establishing who you are and who you're writing to every single time. {{manager_name}} is optional and only worth setting up if you have exactly one recurring audience for these; if you report to different people depending on the document, leave it out and specify the audience manually instead.

Content operations (3)

{{product_name}} = The specific product or feature you write about most
{{newsletter_name}} = The name of your recurring newsletter or content series
{{cta_link}} = https://example.com/your-most-used-call-to-action-link

If you batch a month of content at once, these three alone remove a huge share of the repetitive setup: the product name, the series name, and the link you paste into nearly every piece. {{cta_link}} is worth revisiting whenever a campaign ends, since an expired link quietly sitting in a variable will keep getting pasted into new content until someone notices.

Locale and format (2)

{{language}} = The language you write in most, if it's ever anything but English
{{timezone}} = Your timezone, for anything involving scheduling or "by end of day"

Small, easy to forget, and genuinely annoying to keep retyping when you need them. If you write for one audience but think in another language, {{language}} is worth setting even if you never touch it after the first day, since it's exactly the kind of detail nobody remembers to specify manually until an output comes back wrong.

Contact and links (2)

{{support_email}} = support@example.com
{{calendly_link}} = https://calendly.com/your-actual-booking-link

The two links you paste into the most emails, proposals, and follow-ups. Once saved, they're one {{ }} away instead of a copy-paste from wherever you keep them, and if either one ever changes (a new booking tool, a new support inbox), fixing it in one place beats hunting through old prompts.

That's 20. None of them are complicated, and that's the point: the value here is having them saved at all, not any individual variable being clever. If you're currently keeping this information in a prompt library spreadsheet and copying it in by hand, this is the upgrade path.

Will changing a variable's value break the prompt around it?

One thing worth knowing about how the substitution actually works: your saved value is treated as a current sample, not as the permanent shape of the sentence around it. If you write a prompt using {{industry}} and your saved value happens to be "B2B SaaS," a well-built prompt doesn't quietly lock its surrounding language into SaaS-specific assumptions just because that's what the sample looked like at the time.

That matters the first time you actually change a value. Swap {{industry}} from "B2B SaaS" to "healthcare" and a prompt that was subtly written around the first industry would produce something oddly mismatched. Prompt Architects is built to keep the language around a variable generic enough to survive that swap, which is the entire reason a variable is worth maintaining instead of just editing the prompt text directly each time. If you notice a prompt behaving as though it "remembers" an old value's context after you've changed it, that's a sign the surrounding prompt text itself got too specific, not a problem with the variable.

Do you need a Context instead of a variable?

Variables and contexts solve the same underlying problem, not retyping stable information, at two different sizes. A variable is short: a name, a link, a one-line phrase. A context is long: a full brand voice guide, a product spec, a client's background and history, the kind of material that runs to paragraphs.

If you catch yourself trying to cram several sentences into a variable's value, that's the signal you actually want a context instead. Keep variables atomic. It's what keeps {{company_name}} readable inline in a sentence, rather than turning every prompt into a wall of pasted background material. A rough rule that holds up in practice: if you'd be comfortable saying the value out loud in one breath, it's a variable; if you'd need to read it off a document, it's a context.

Can your team share global variables?

On the Team plan, a variable can be marked shared so every teammate's prompts pull the same value. Define {{company_name}} or {{support_email}} once at the team level, and every member's future prompts use it. Update it once, and the change reaches the whole workspace without anyone editing an individual prompt. This is the same problem the team prompt playbook covers at the library level, just applied to the smaller, atomic facts a shared library alone doesn't standardize. Team pricing is $10/month plus $3.50 per teammate (or $100/year plus $35 per teammate), for 2 to 20 members, with the owner's seat free.

Each member also keeps a personal set of variables that stays private and never shows up in the shared workspace. That's useful for anything account-specific that shouldn't be standardized across the team, like a personal writing quirk or a variable you're still testing before proposing it as a team default. A new hire inherits every shared team variable the moment they join the workspace, which is a small thing but a real one: they start writing prompts with the company's actual name and tone already in place, instead of learning it from a wiki page first.

How many variables can you save?

This is genuinely plan-dependent, and worth checking directly rather than guessing: how many variables you can save isn't listed on the public pricing comparison table the way prompt history or library size are, and it varies by plan. Open the Variables page in your own account to see your current count and limit rather than assuming a number from anywhere else, including this post.

Twenty variables is a starting list, not a ceiling. Once the boring, stable facts about your work are saved, the actual gain shows up quietly: every prompt you write afterward starts a little further along, because the parts that never change stopped needing to be retyped.