Shortwave Pricing Explained: Plans, AI Limits and Caps

Shortwave pricing explained, in short: the tiers are separated less by features than by how much AI you get and how far back it can look. Each step up raises the model class, the daily AI request quota, the years of searchable history, the threads a search may read, and the number of AI filters you can run.

You will not find a dollar figure on this page. That is deliberate, and the reason is the reason this page exists.

What follows is the part that does not expire: what each tier is for, how AI usage gets metered in this category, what happens when you hit a cap, and how to size your own volume before you buy.

Why this page will not print a price#

Researching this cluster, a single search for Shortwave's pricing produced three mutually contradictory price sets. They did not differ by a few dollars — they disagreed about how many tiers exist.

It goes further than third parties. On 29 July 2026, Shortwave's pricing page listed three paid plans, while its own billing docs on the same domain still described a line-up including Free and Pro. Docs lag pricing pages everywhere, ours included.

But it means a number copied from a review, or from the wrong page on the vendor's own site, has a real chance of describing a plan you cannot buy. A confidently wrong price is worse than none, because it reads like an answer.

What the tiers are actually for#

Most software tiers gate features: the cheap plan lacks the reporting, the expensive one adds SSO. Shortwave's tiers are mostly not shaped that way.

Across the paid plans the assistant is the same product. What changes is how capable the model is, how often you may call it, and how much mail it may read — which is why the cheapest plan can be the wrong plan.

The free plan has narrowed#

Shortwave used to offer a permanently free tier. As of July 2026 that has changed: the pricing page we loaded showed paid plans and a 14-day trial, with no permanently free option presented on it.

This is a normal business decision, not a betrayal. AI inference costs real money per request, and most products in this category have tightened their free tiers for the same reason.

It matters to you in one way: the free evaluation path older articles describe may no longer exist, so confirm the current trial terms before you plan a rollout around them.

How AI limits actually work in this category#

This part is worth understanding whichever product you buy. A cap on messages is not a cap on AI actions, and neither is a cap on model quality — knowing which one you are buying tells you what breaks first.

The ladder as published on 29 July 2026#

Here is the escalation, tier by tier — limit dimensions, not prices, date-stamped because this detail goes stale quietly. Weighing the business tier against the premier one? This table is the whole decision.

Common misconceptions#

• That the plans differ by features. Mostly they differ by AI allowance and search reach.
• That a large daily quota is generous. Every summary, search and redraft spends it.
• That unlimited history means unlimited search. Depth per search is capped separately, at every tier.
• That the free plan you read about still exists. It narrowed, and the page we loaded led with a trial.
• That Shortwave is a plugin. It is a full Gmail client, not an overlay — a real point in its favour.

Estimate your own AI volume before you buy#

Two minutes of arithmetic beats a month of regret. Do this for any product in this category.

1. 1

   Count your real daily volume

   Search your mailbox for the last seven days of received mail and divide by seven. Use a busy week.

2. 2

   Decide what you would hand to AI

   Most people triage everything but only draft with AI on a fraction of threads. Multiply by that fraction.

3. 3

   Add the invisible requests

   Summaries, searches, rewrites and filter runs all spend quota. Budget two to three requests per thread you touch.

4. 4

   Check your oldest thread that still matters

   If the oldest message you needed this year sits outside a tier's search window, that alone decides your tier.

5. 5

   Date-stamp what you find

   Screenshot the pricing page with the date and circulate both. Anyone comparing it to an article later needs to know which is stale.

What we charge, and where Shortwave still wins#

Full disclosure: we build AI Emaily, so treat this as an interested party's summary with the limits stated. We publish our own numbers because we can verify them — the same standard that stops us guessing at someone else's.

AI Emaily has three authority modes: Manual, Copilot and Autopilot. In Copilot nothing sends without approval, every agent action is undoable, and the agent's work lands in an audit log. Gmail, Outlook and IMAP sit in one inbox, and drafting voice comes from a Personal Context you write plus client profiles you set. We do not train on your mail.

• Our limits: the desktop app is downloadable on macOS (Apple Silicon only) and Windows, built as an Electron shell rather than a native binary. No Linux build. Android is a PWA today.
• Where Shortwave is ahead, it is ahead: if you live entirely in Gmail and want the deepest semantic search over your own archive, that is what it was built to do.
• One thing to weigh either way: its founding team has publicly shifted focus toward a separate agent product. That is continuity risk to price in, not a shutdown.

The short answer, once more#

Shortwave charges a per-person rate across tiers that escalate on AI capability, not features — model class, daily quota, searchable years, threads per search, filter count. The free plan has narrowed, and the current line-up no longer matches most articles about it.

The figure belongs to Shortwave's pricing page, read today, with the date beside it.