How to Import a ChatGPT Workout Plan Into a Tracker App

By the Trunk teamUpdated September 7, 202613 min readPrices verified September 7, 2026

Three routes get a ChatGPT workout plan into a tracker without re-typing it: HevyGPT saves routines into Hevy from inside ChatGPT, Ellim's text import turns a pasted plan into a trackable routine for free, and Trunk hands any assistant a link that builds the program in the app. Liftin' has a fourth route in beta with no public setup documentation. On Strong and Boostcamp there is no import path at all, so you rebuild the plan by hand.

Three routes that import a ChatGPT workout plan without re-typing

Stated before you read the table: Trunk is ours. It appears on the same terms as everything else in the list, and each figure is linked to the page it was taken from.

Prices in USD, verified 2026-09-07 on the linked US App Store and vendor pages. Apple's prices in other countries are not currency conversions, so this comparison holds in the US only. Prices change without notice.

Step one: make ChatGPT write something importable

Ask for a program in normal conversation and you get prose — paragraphs of coaching notes with the sets and reps buried inside them. Prose is what forces the re-typing. Tell the model to output a table instead and every route below gets easier, manual entry included.

Paste this and fill in the brackets:

Build me a [4-day upper/lower] program for the next 8 weeks.

Output rules:
- No prose. Output one table and nothing else.
- Columns: day, exercise, sets, reps, target RPE, notes
- One row per exercise. Name each day exactly once (e.g. "Upper A").
- Full exercise names, no abbreviations.
- Maximum 6 exercises per day.

About me: [age, years training, days per week, equipment I have,
my best recent sets on the main lifts, injuries to work around].

A table pastes cleanly into Ellim's importer, reads correctly to HevyGPT, and is far quicker to re-key into Strong than a wall of text. Keep the chat open afterwards — the first thing you will want is a change to week two.

Hevy: HevyGPT writes the routine for you

Hevy has the most direct path, and it is the bigger, more mature product: 4.9 stars from roughly 91,000 US App Store ratings as of 2026-09-07, with Android, a web app, and an Apple Watch app that Trunk does not have.

HevyGPT is an official custom GPT inside ChatGPT. Connect it from Hevy → Profile → Settings → Integrations, describe the program you want in ChatGPT, and ask it to save the routines to your Hevy profile. ChatGPT can also read your training history through the same integration, so the block it writes can be based on what you actually lifted.

Three limits matter before you rely on it, all stated on Hevy's own feature page (source):

• HevyGPT does not require Hevy Pro. What it requires is room: free accounts are limited to four routines, and a plan with more than four unique workouts cannot be imported. The cap is account-wide, so routines you already have count against it. A 3-day push/pull/legs fits inside it. A 5-day upper/lower/push/pull/legs, PHUL, or Upper A/B plus Lower A/B plus an accessory day does not.
• ChatGPT only. No Claude or Gemini equivalent of HevyGPT exists.
• It does not work in the Android ChatGPT app, so Android users have to run ChatGPT in a browser.

Hevy does not state which ChatGPT tiers can run a custom GPT, and we could not verify whether a free ChatGPT account can. If you are deciding between the two integrations specifically, we put them side by side in HevyGPT vs Trunk Connect AI.

Hevy's iOS v3.1.12 release note, dated 2026-09-03, says you can now send workouts to ChatGPT and Claude from the finish-workout screen. No help article describes it, so the mechanism, the direction it runs, and whether it needs Pro are all undocumented.

Ellim, Boostcamp and Strong: paste it, build it, or type it

Ellim — free text import

Ellim's routine import is the closest thing to a universal paste box in this group. Photograph a plan or paste it as text, and Ellim's AI parses the exercises, sets and reps into a trackable routine.

The import sits in the free column on Ellim's own site, and that free tier carries no routine cap and includes a 3,500+ exercise library with demonstrations — the thing Trunk does not have at all.

What Ellim will not do is let your assistant read your log. It exposes no API, no MCP server, and no bring-your-own-assistant option; its other AI features are Ellim's own, at $17.99/month or $99.99/year as of 2026-09-07, with no one-time tier (source). Data out is a JSON export you can download once every 24 hours and upload to an assistant yourself. Ellim's own site says iPhone only, no Android yet.

Boostcamp — rebuild it in the free builder

Boostcamp's answer to "where does my program come from" is usually not an AI. The free tier carries 11,000+ programs, including most of its 130+ coach-designed routines from named coaches, at 4.8 stars from 10K US App Store ratings. Read those before you write a prompt at all.

There is no import path either way, and Boostcamp's own comparison page concedes it does not offer a one-click import even from Strong. The custom program builder is free with no cap on how many programs you create, so re-keying a ChatGPT plan costs time and nothing else. Boostcamp's own AI is a Pro-only questionnaire that produces a multi-week block — $14.99/month or $59.99/year, verified 2026-09-07 (source) — rather than a chat you can hand a plan to.

Evidence that people already re-key by hand: a community program on Boostcamp is titled, verbatim, "Chatgpt beginner's workout plan".

Strong — type it in, three routines at a time

Strong is an excellent logger with 4.9 stars from about 109,000 US App Store ratings, refined since 2011, with a standalone Apple Watch app you can log a whole session on without touching your phone. It also has no AI and no import of any kind.

Its help center states that exported files cannot be imported back into Strong (source), so the only route is reading the plan and building templates by hand. The free tier stops at 3 custom routines (source), so a 4-day split means Strong PRO.

Trunk: one link, any assistant, structure only

Trunk+ is a $9.99 one-time in-app purchase. No subscription tier exists in the app. It turns on Connect AI, which runs like this:

1. Settings → Trunk+ → Connect AI → Generate Key. Trunk creates a random token and builds a ready-made prompt around it.
2. Copy Prompt, then paste it into ChatGPT, Claude, Gemini, or any assistant that can open a web link. The prompt carries a private URL to your logged sets — reps, weight, volume, per-set notes, dates, plate configuration — which the assistant opens and reads. Parameters such as ?exercise=bench or ?last=5 narrow the view.
3. Ask it to build a program. It answers with a single import link. Tap the link, check the confirmation page listing exactly what will be created, and tap Add to Trunk.
4. Revoke Key kills the link whenever you want, and generating a new key invalidates the old one.

The AI gets three write actions, all deliberately narrow: create a program, add days to a program you already have, and add exercises to an existing day. It will not overwrite a program that already exists.

The import carries structure only. Sets, reps and weights are never written to your log — the AI plans, you lift, and every number in your history is one you actually did.

Each assistant behaves differently:

• ChatGPT sometimes answers from memory instead of opening a link, so the generated prompt includes an explicit block that forces it to fetch the page and list the programs it found before answering.
• Claude needs web search switched on for that chat, from the "+" button in the composer. That toggle is documented for the web app; Anthropic does not document the iOS composer. Anthropic's help center says Claude can retrieve full content from URLs you provide, and warns that the whole page lands in the context window and counts against your limits (source). That is what ?last=5 is for.
• Gemini: Trunk's developer reports that Gemini opens the link. Google publishes no page confirming that Gemini reads a pasted URL, and we have not independently reproduced it.

The full step-by-step lives in connecting ChatGPT to a workout tracker, and the Claude version in using Claude to write your program.

Worth being blunt about the link itself: it doubles as your credential. There is no password sitting behind it, so whoever you forward it to can read every set you have ever logged. Revoking is the remedy — generating a fresh key invalidates the old one on the spot, which makes a link you regret sending a thirty-second problem rather than a permanent one.

Liftin': the other bring-your-own-assistant route

Trunk is not the only tracker that will talk to the assistant you already use. Liftin' announced "Liftin' MCP" on 2026-07-22 through its official Mastodon account: a beta billed as full integration with ChatGPT or Claude, for analyzing your history, changing routines and creating a program in natural language (source).

Three things keep it out of the step-by-step above. Setup instructions live only in a Reddit post we could not fetch. Gemini is not named. And whether the beta requires Liftin' Unlimited is unverified — the free tier caps you at 5 tracked workouts a month, and Unlimited is $2.99/month, $24.99/year or $99.99 one-time as of 2026-09-07 (source). Liftin' is iOS-only, rated 4.7 from 745 US App Store ratings, with a standalone Apple Watch app.

What the paid tiers cost

The plan itself is free to generate: ChatGPT, Claude and Gemini all have $0 tiers as of 2026-09-07. What costs money is the app that logs it.

USD, verified 2026-09-07 from the linked US App Store and vendor pages; prices change without notice, and Apple's prices in other countries are not currency conversions. Strong's US listing also shows lower-priced SKUs ($23.99/yr and $79.99 one-time), and which SKU a new buyer is served is not stated anywhere. Hevy's listing shows a second "Hevy Pro - Monthly" SKU at $3.99 with no published explanation. Boostcamp advertises its annual plan as "$4.99/mo", billed $59.99 yearly.

Read that table honestly and the cheapest route to an imported plan is $0: free Hevy plus a ChatGPT account, or Ellim's free text import. Trunk+ costs $9.99 once and buys one thing those do not have — any assistant instead of one. On routine caps it ties Ellim and Boostcamp, both uncapped on free, and beats only Hevy's 4 and Strong's 3.

Paying by the month is the expensive way to do any of this. Twelve months of Hevy Pro at $2.99 is $35.88 against $23.99 for the year; twelve months of Ellim Plus at $17.99 is $215.88 against $99.99. One-time pricing is not unique to Trunk either — Hevy Pro Lifetime is $74.99, Strong PRO Forever is $99.99, and Liftin' Unlimited Lifetime is $99.99. Trunk's $9.99 is the cheapest one-time unlock among the iOS trackers we researched; on Android, FitNotes charges $6.99 once for its Supporter app and has no AI at all.

Who should pick which

• You already use Hevy and live in ChatGPT. Stay put. HevyGPT is bidirectional inside the chat with no copy-paste, it needs no Hevy Pro to start, and Hevy is a better-established app on more platforms. Budget $23.99/yr or $74.99 once if your split runs past four routines.
• You want the plan in an app in under a minute and nothing more. Ellim's free text import. Paste, confirm, log.
• You are settled on Strong or Boostcamp. Re-type it from the table your prompt produced. Check Strong's 3-routine free cap first, because that is the limit you will hit, not the typing.
• You use Claude or Gemini, or you want the assistant reading your whole history rather than one workout. Trunk+ at $9.99 once, on iOS — or Liftin's MCP beta if you only need ChatGPT or Claude and are willing to hunt down setup instructions. If you need Android, an Apple Watch app, or an exercise library with video demonstrations, pick something else; Trunk has none of those.
• You refuse to rent software. Compare $9.99 for Trunk+, $74.99 for Hevy Pro Lifetime, and $99.99 each for Strong PRO Forever and Liftin' Unlimited Lifetime, then pick on features rather than on the word "lifetime".

Whichever you choose, the method is the same: make the model emit a table, get the structure into the app once, and log every set yourself. A plan the AI wrote is a hypothesis. Your log is the only thing that tells you whether it was right.