How to connect Instagram to Claude and read the organic side

Reading Instagram inside Claude solves one specific problem: it stops you from judging content by feel. When profile metrics and campaign metrics live in the same conversation, "does this format work?" stops being an opinion and becomes a comparison.

This article covers what you need to connect, what the API actually returns, and the questions that become worth asking.

What you need first

A business or creator account. Personal accounts have no metrics API. Converting is free and takes two minutes in the app settings.

A link to a Facebook Page. The Instagram API runs through Meta's infrastructure, and the link is what grants access. If the Page exists but is not linked, the connection fails without a clear message — the number one cause of trouble at this step.

Admin permission. You need to administer the Page, not just the profile. On client accounts that usually comes through the business portfolio.

A connector. It is what bridges the API and the assistant. The mechanism is in what is MCP.

How the connection works

The flow is the same as any OAuth authorization and takes under a minute:

1. In the connector, choose to connect with Facebook.
2. You go to Meta's page and sign in with your credentials, there.
3. The screen lists the Pages and linked Instagram accounts. Select the ones you want to grant.
4. Approve. The connector receives a read-scoped token.
5. In Claude, enable the connector in settings and start asking.

Your password never passes through the connector, and access is revocable at any time in Facebook's app settings. What that implies is in AI data security in marketing.

What the API returns

What it does not return, and this matters before you build expectations: data from profiles that are not yours, bulk comments for sentiment analysis, and unlimited history. Stories in particular have a short availability window — read them late and they are gone.

The questions that become worth asking

With organic and paid in one conversation, some questions become possible for the first time:

• "Which posts from the last 30 days got the most saves?"
• "Compare this month's organic reach with last month's."
• "Which format held attention better: Reels or carousel?"
• "Did follower growth track the increase in ad spend?"
• "Are the themes that perform organically the same ones that perform in paid?"

The last one changes creative decisions the most. Organic content is a cheap laboratory: if a theme holds attention with no money behind it, it has a real shot as an ad. Testing organically costs time; testing straight in paid costs budget.

The mistake of comparing organic and paid in the same column

Organic reach and paid reach are not the same metric with different origins. Organic is distribution conditioned on demonstrated interest; paid is distribution bought for an audience you defined.

Adding them into a total reach number inflates the result and informs nothing. Putting them side by side, in separate columns, informs a great deal: it shows whether the content sustains attention without money behind it.

The same caution applies to engagement. Likes and comments are vanity metrics when they appear alone in a client report, and become useful information when compared across formats in the same period.

A twenty-minute weekly routine

Turning the connector on is the easy part. What produces results is having three or four fixed questions you repeat weekly, always with the same date range.

Monday, ten minutes. "Compare reach, saves and profile visits for the last 7 days against the previous 7." The question is about trend, not a snapshot.

Monday, five more minutes. "Which three posts in the period got the most saves and shares?" Saves and shares correlate with attention-holding content far better than likes do.

Monday, last five minutes. "Did any theme appear in the top three posts this month?" This is the question that becomes a creative brief for paid.

Once a month. "Compare follower growth against ad spend in the same period." It answers whether the investment is building a base or only buying temporary reach.

Twenty minutes a week, always the same questions. The repetition is what builds the history that makes comparison meaningful — and it is exactly what gets lost when every week's analysis starts from scratch with a different cut.

Which organic metrics actually predict paid performance

Not every column is worth carrying into the creative decision. Three earn their place, and the rest are context.

Saves. The strongest signal in the set. A save means the person intends to come back to it, which is as close to intent as organic gets. A post with modest reach and high saves is a better ad candidate than a post with huge reach and none.

Shares. Second strongest, and it measures something slightly different: whether the message is worth attaching your name to. Shares predict how a creative behaves in cold audiences better than any engagement rate.

Watch time on Reels, not plays. Plays count the autoplay; watch time counts the attention. Ranking your Reels by average watch time usually produces a completely different order than ranking by views — and the watch-time order is the one that transfers to paid.

What to leave out of the decision: follower count changes over short periods, likes in isolation, and comment counts when a giveaway is running. All three move for reasons that have nothing to do with whether the message works.

If the connection fails

Three causes account for most cases, in this order:

1. The account is still personal. The symptom is the account not appearing in the selection list.
2. Instagram is not linked to the Page. The symptom is the Page showing and Instagram not.
3. Insufficient Page permission. You are an editor, not an admin. The symptom is a generic permission error after approving.

A fourth cause, rarer and more irritating: the Instagram account is linked to a Page that was archived or recently moved between portfolios. There the authorization passes, the token is issued, and metrics come back empty. The quick test is opening Business Manager and confirming the Page is active in the portfolio you are authorizing from.

With those cleared, the connection usually works on the first try. And the outcome it unlocks is the most direct one there is: no more arguing about whether the content worked — just look. The document that holds both readings is on the paid media reporting page.