Turn your Instagram Insights into a 30-day plan with Claude
Hey, it's Cindy 🌱 You commented GROWTH, so here is the full setup: a live connector that pipes your Instagram numbers into Claude on an ongoing basis, and a two-minute manual export that costs nothing.
btw the waitlist for AI System Club is now open! If you're keen on using AI to automate your content, video editing, job search, business or build AI agents, feel free to sign up (only if you want to hehe). Ps. the waitlist gets first dibs:)
the truth first
There is no secret Instagram AI
Instagram has not released an AI that reads your account and hands you a content plan. What the viral reels are doing is connecting a third-party data pipeline, or pasting their own screenshots in, and then gatekeeping the prompt. Both routes are legitimate. Neither is magic. Claude cannot see the algorithm's internals or anyone's private numbers, so what you get is a genuinely sharp read of your data, which is more than most people ever do.
Route A, the live connector Windsor.ai authorises your Instagram account once and keeps syncing. Claude can query fresh numbers any time you ask. More setup, no re-exporting.
Route B, the manual export You export or screenshot your Insights and paste them in. Zero signups, zero third parties, works on any plan. You redo it each time.
🌱 Not sure which? Do route B this afternoon to see whether the output is actually useful to you. If you find yourself repeating it every fortnight, that is your signal to set up route A properly.
route a
Connect Instagram to Claude with Windsor.ai
Windsor.ai is a marketing data platform that publishes a connector for Claude. It authorises your Instagram account on its side, then exposes that data to Claude over a live connection, so you ask questions in chat instead of exporting spreadsheets.
📱 Make sure your account qualifies. The Instagram Insights connector needs a Business or Creator account linked to a Facebook Page, and you need admin access to it. On a personal account, open Instagram settings, find account type and tools, and switch to a professional account. It is free and reversible.
👩💻 Authorise the data source. Go to onboard.windsor.ai, create an account, and pick Instagram Insights as your data source. You authenticate through Facebook and then choose exactly which Instagram profile to connect.
🔗 Add the connector inside Claude. Open Customize, then Connectors, and use the + button to add it. You can also reach it mid-chat: click the + button or type /, hover Connectors, then choose Manage connectors. Sign in to Windsor when it prompts you.
✅ Set the permission once. When Claude first calls the connector it asks for approval. Choose the always-allow option so you are not re-authorising it on every single question.
🔎 Sanity-check the data before you trust it. Ask something you already know the answer to, for example your reach last month or your top three posts. If the numbers match what Instagram shows you, the pipe is working. If they do not, fix that before you build a plan on top of it.
💡 Two different Instagram connectors, pick deliberately.Instagram Insights is your own account: saves, shares, reach, story taps and exits, follower demographics, profile visits, roughly a hundred fields, and it needs admin access. Instagram Public Data needs no account access at all and reads any public profile, but only externally visible things like follower counts, captions and post engagement. For a content plan you want Insights. For competitor research you want Public Data.
The plan gate, stated honestly. Windsor.ai runs a 30-day free trial (1 user, 10 data sources, 15 accounts, 30 days of history), and when that ends there is a forever-free plan limited to 1 data source and 1 account, which is exactly enough for one Instagram profile. Paid plans start at $23 per month, or $19 billed annually (current pricing). On Claude's side, connectors work on every plan including Free, though free accounts are capped at one custom connector (connectors, Claude help centre). Also worth knowing: this route sends your Instagram data through a third party, and it covers organic analytics only, not ads.
route b
Or export it yourself in two minutes
No signups, no third party, no subscription. This is what I would do first.
💻 Desktop, the better version. Open Meta Business Suite at business.facebook.com, go to Insights, set your date range to the last 90 days, and hit Export to download the CSV. This gives you real post-level data, which is what makes the analysis specific rather than vague.
📱 Phone, the fast version. Open your Professional Dashboard, go to Account Insights, pick 30 or 90 days, and screenshot the content-performance list sorted by reach and again by saves. There is no CSV on mobile, and Claude reads screenshots fine.
📋 Give it context, not just numbers. Along with the file, tell Claude who you are, who you are trying to reach, and what you are actually selling or building towards. Identical data with different goals produces different plans, and this single paragraph is the difference between a useful calendar and a generic one.
copy these
The 3 prompts that do the work
Run them in order, in the same conversation, so each one builds on the last.
📊 1. the honest audit
You are my Instagram strategist. Here is my account data [attach the export, paste the screenshots, or pull it from the Windsor connector]. My audience is [who], and what I am building towards is [goal].
Analyse this properly and be specific rather than encouraging. Work out an engagement rate per post and rank every post against my own median, so I can see what genuinely over-performed instead of what merely got the most likes. Then group the posts by format and by topic and tell me which combinations actually travel beyond my existing followers, using reach from non-followers where that data exists. Finish by naming the three things I am clearly wasting effort on, and be explicit about which of your conclusions are well supported by the data versus which are educated guesses.
🧠 2. find the pattern, not the post
Now look only at my top ten posts by the ranking you just built, and work out what they have in common at a structural level rather than a topical one.
Compare their opening lines and first three seconds, their length, their format, the emotion they lead with, whether they teach, argue or show something, and what they ask the viewer to do at the end. Do the same for my bottom ten so we have a contrast, and tell me honestly where the two groups overlap, because that is where the pattern is weaker than it looks. Give me the result as a short list of repeatable rules I could hand to someone else, each one with the specific posts that support it.
🌱 3. build the 30-day plan
Using the rules you just derived from my own numbers, build me a 30-day content calendar. Do not give me generic content ideas, every single slot has to trace back to something in my data.
For each post give me the date, the format, the specific topic, the actual opening line, and one sentence explaining which finding from my data it is based on. Weight the mix towards the formats and topics that over-performed, but include three deliberate experiments in areas I have never tested, and label them as experiments so I know not to judge them on the same benchmark. Keep the volume realistic for one person, and at the end list the three metrics I should check at the halfway point to know whether the plan is working.
be realistic
What this can and cannot tell you
It reads patterns, not the algorithm. Nothing here gives you visibility into ranking signals. It gives you a disciplined read of your own results, which is genuinely valuable and completely different from insider knowledge. Any tool promising the latter is selling you something.
Small accounts get noisier answers. Under roughly 30 posts in the date range, one viral outlier can drag every conclusion sideways. Ask Claude to flag outliers and re-run the ranking with them excluded, then compare the two versions before you commit to a plan.
🌱 Re-run it, do not set and forget. The point of the live connector route is that this becomes a monthly habit rather than a one-off project. Once a month, run prompt one again and ask what changed since last time. That comparison is where the real compounding is.