Instagram MCP: setup guide and options
An Instagram MCP server connects your agent to Instagram data: profiles, posts, reels, comments and hashtag search. This guide covers your options, setup in Claude, ChatGPT and Cursor, and when a hosted alternative fits better.
An Instagram MCP server connects an AI agent such as Claude, ChatGPT or Cursor to Instagram data: profiles, posts, reels, comments and hashtag search. The quickest route is a hosted server. Manifold is hosted, so there is no Instagram login, Meta developer app or API key of your own to set up. You pay per call, and every workspace starts with 500 free credits.
This guide explains what an Instagram MCP server does, compares your options, walks through setup in Claude, ChatGPT and Cursor, and lists the Instagram tools, example prompts and common questions.
What is an Instagram MCP server?
MCP (Model Context Protocol) is the open standard AI clients use to call outside tools. An Instagram MCP server exposes Instagram as a set of tools your agent can call: get a profile, list posts or reels, search posts by hashtag, read comments and pull a reel's transcript. You ask a question in plain words, the agent picks the calls, reads the results and replies with a summary.
People use an Instagram MCP server to:
- Research competitor content: what they post, how often and which posts do best
- Size up a creator or brand before a partnership or outreach
- Track posts under a hashtag or a branded tag
- Mine comments for sentiment, questions and complaints
- Pull a reel's transcript without watching it
Your options: open source, hosted or Meta's Graph API
There are three kinds of Instagram tool. Pick by what you want to reach and whether you want to manage access yourself.
Open-source servers
Most GitHub results for "Instagram MCP server" are self-hosted servers you run yourself. Some are built on Meta's official Instagram API for Business and Creator accounts you manage. Others log in to Instagram's private API with a username and password, which reads any public account but risks restrictions. A few examples:
- William-Gao/instagram-mcp (Python, MIT). Uses Meta's Instagram API with Instagram Login, so no Facebook Page is needed for the core tools. More than 25 tools for profile, media, publishing, comments, insights and DMs. Add a Facebook Page token and it can also look up other Business or Creator accounts and search hashtags. DMs need Meta's App Review.
- adelaidasofia/instagram-mcp (Python, FastMCP, MIT). Official Instagram Graph API for Business or Creator accounts linked to a Facebook Page. 29 tools for accounts, media, publishing, insights, comments, hashtag and account discovery, and DMs that need App Review. Runs many accounts from one server.
- anand-kamble/mcp-instagram (TypeScript, MIT). Logs in with your Instagram username and password. Searches accounts, reads profiles, posts, stories, comments and your home feed, and can like, comment and follow. Its README warns that your account can be restricted or suspended.
A local setup in Claude Desktop is a few lines in claude_desktop_config.json. For anand-kamble/mcp-instagram it looks like this:
{
"mcpServers": {
"mcp-instagram": {
"command": "npx",
"args": ["-y", "mcp-instagram"],
"env": {
"IG_USERNAME": "your_instagram_username",
"IG_PASSWORD": "your_instagram_password"
}
}
}
}Self-hosting is free and gives you full control, with trade-offs:
- Meta's API is built around your own accounts. It works on Instagram Business or Creator accounts you manage, through a Meta app you register. Looking up other accounts or hashtags needs the Facebook Graph API path, which means a linked Facebook Page, and it only covers Business or Creator accounts.
- Unofficial APIs risk restrictions. Servers that log in with your username and password drive Instagram's private APIs. Instagram's Terms of Use prohibit this, and accounts doing it can be restricted or banned.
- Local servers stay local. A server that runs over stdio works in desktop clients on your machine, not in Claude on the web, on your phone or in ChatGPT, unless you host it behind a public URL yourself.
- You maintain it. Updates and breakages when Instagram changes are yours to handle.
Hosted servers for publishing
A few hosted servers use Meta's official API to publish posts, Reels and Stories on your own Instagram Business or Creator account:
- Blotato at
https://mcp.blotato.com/mcp. Publishing, scheduling, analytics, comments and DMs on Instagram, plus eight other platforms. Starter is $29/mo with 20 social accounts and a 7-day free trial. - Outstand at
https://mcp.outstand.so/mcp. Publishes photos, Reels, carousels and Stories, reads post and account metrics, and replies to comments on posts it published. Connects through Meta's OAuth. $19/mo with 3,000 posts, across 12 platforms.
These servers let you publish to your own account from your agent. They do not research public accounts you do not own.
Hosted servers for research
A hosted server runs in the cloud. You paste one URL into your client, and nothing runs on your machine. A few read public Instagram data:
- HasData at
https://mcp.hasdata.com/mcp?apis=instagram, with a HasData API key in a header. Three tools: a public profile, an account's post feed, and the comments on one post. 10 credits a call, with 1,000 free credits a month (100 calls) and paid plans from $59/mo. - Instapi at
https://instapi.co/mcp, with an Instapi API key. Three data tools: search users, hashtags and places, get a profile, and list an account's recent posts, with images and videos described as text. 10 free credits, then $1 per 100 credits, at 1 credit a call. - Manifold reads public profiles, posts, reels, comments, hashtags and reel transcripts. There is no Instagram login, no Meta app to register and no API keys of your own. You pay per call: 500 free credits per workspace to start, then top-ups from $10 or Pro from $20/mo. It works in Claude on the web, desktop and mobile, ChatGPT, Cursor, Claude Code and any other client that supports remote MCP servers. The same server also covers SEO, AI answers, Reddit, other social platforms, ad libraries and leads, so your agent can check Instagram and, say, what your competitors post on TikTok in one conversation.
Meta's official ads MCP
Meta publishes an ads MCP server at https://mcp.facebook.com/ads, documented at developers.facebook.com. This server manages your own Meta ad account: reporting, campaigns, catalogs and tests. Its documentation lists no tools for reading public Instagram profiles or posts, so for Instagram content research you need a third-party server.
Compared
| Open-source, official API | Open-source, unofficial | Hosted publishers | Manifold (hosted research) | |
|---|---|---|---|---|
| Setup | Install locally, register a Meta app, get an access token | Install locally, give it your Instagram login | Paste one URL and sign in | Paste one URL and sign in |
| Uses your Instagram account | Yes, a Business or Creator account through your Meta app | Yes, your username and password | Yes, connected through Meta's OAuth | No |
| Whose data it reads | Your own accounts, plus other Business or Creator accounts with a Facebook Page token | Any public account | Your own accounts | Any public account |
| Publishing, comments and DMs | Yes (DMs need Meta's App Review) | Comments, likes and follows | Yes, that is what they are for | No, your agent drafts and you post |
| Reel transcripts | No | No | No | Yes |
| Clients | Desktop clients on your machine | Desktop clients on your machine | Claude (web, desktop, mobile), ChatGPT, Cursor, Claude Code and other remote MCP clients | Claude (web, desktop, mobile), ChatGPT, Cursor, Claude Code and other remote MCP clients |
| Cost | Free software, plus your own time | Free software, plus your own time | $29/mo (Blotato), $19/mo (Outstand) | 500 free credits per workspace, then pay per call |
| Other data in the same server | Instagram only | Instagram only | Publishing across platforms | SEO, AI answers, Reddit, other social, ad libraries and leads |
| Best for | Managing your own Instagram Business or Creator account | Developers who accept the account risk | Social media managers who post and schedule | Anyone who needs public Instagram data in their agent, in any client |
Set up an Instagram MCP server with Manifold
You need a Manifold account, which your client creates for you the first time you sign in. The server URL is the same in every client:
https://mcp.manifoldmcp.com/mcpClaude
- In Claude on the web or in the desktop app, open Settings → Connectors.
- Choose Add custom connector. Name it Manifold and paste the server URL.
- Choose Connect, then sign in to Manifold in the window that opens and approve Claude.
- Start a chat, check that Manifold is on in the chat's tools menu, and ask an Instagram question.
A connector you add on the web is also there in the desktop app and on your phone. In Claude Code, run claude mcp add --transport http manifold https://mcp.manifoldmcp.com/mcp, then /mcp to sign in. Full steps are in the Claude guide and the Claude Code guide.
ChatGPT
- In ChatGPT on the web, open Settings → Security and login and turn on Developer mode. ChatGPT needs it to add an MCP server by URL.
- Go to chatgpt.com/plugins, click Add at the top right, then choose Create MCP app.
- Name it Manifold and paste the server URL under Connection. Leave Authentication on OAuth.
- Tick I understand and want to continue, then choose Create. Sign in to Manifold when ChatGPT asks.
- In a chat, type @Manifold or pick it from + → More, and ask an Instagram question.
On a Business workspace only an admin can use developer mode, so ask them to add Manifold under Workspace settings → Apps → Create. See the ChatGPT guide.
Cursor
- Open Cursor Settings → MCP, add a server and paste the URL, or use Add to Cursor on the Cursor guide.
- Cursor lists Manifold. Choose it to sign in, and approve Cursor in the browser.
- Ask an Instagram question in the agent chat.
To share the server with everyone on a project, add it to .cursor/mcp.json:
{
"mcpServers": {
"manifold": { "url": "https://mcp.manifoldmcp.com/mcp" }
}
}Other clients
Codex, Gemini CLI, Grok and other clients are covered in the client guides.
What you can do with Manifold's Instagram tools
Your agent picks the tools for you, but knowing what each does helps you ask better questions.
| Tool | What it does |
|---|---|
instagram_get_profile | One account: followers, following, post count, bio, verified, website and whether it is a business account |
instagram_get_posts | The posts on one account, newest first, with likes, comments, shares and views where Instagram publishes them |
instagram_get_reels | The reels on one account, newest first, with the same engagement numbers |
instagram_get_post | One post or reel in full: caption, engagement, media, length and whether it is an ad |
instagram_get_comments | The comments on one post or reel, with their likes, and replies when you ask for them |
instagram_search_posts | Posts under one hashtag, ranked by Instagram, from the past day up to all time |
instagram_get_transcript | What is said out loud in one reel up to two minutes long, as text |
A few things worth knowing:
- Hashtag search only. Instagram does not offer free-text search across all posts, so
instagram_search_postssearches by hashtag. If Instagram has no way to answer a question, there is no tool for it. - Repeat calls cost less. Most results are cached for hours, and a repeat of a call your workspace already paid for is free until the cache expires.
- You see the price first. Every result reports what it cost,
max_creditscaps what a call can spend, anddry_runprices a call for free. - Your agent does the reading. It finds and reads accounts, posts and comments, reports back and drafts captions or replies if you ask. You decide what to post.
For parameters and result fields, see the Instagram tools docs.
Example prompts
Paste any of these into Claude, ChatGPT or Cursor once Manifold is connected.
Creator research. "Compare these five creators on follower count, post cadence, average engagement and whether they are business accounts. [handles]"
Hashtag monitoring. "Show me the top 20 posts under #veganrecipes from the last month and which accounts posted them."
Content analysis. "What are the ten most-liked reels on this brand's account this quarter, and what do they have in common? [handle]"
Comment mining. "Read the comments on our competitor's latest launch post and group them by sentiment, with quotes and links."
Reel transcripts. "Pull the transcripts of these five reels and list the claims they make. [reel URLs]"
Competitor tracking. "Which of our competitors posted in the last week, and what did they post about? [handles]"
Run this in your agent
Connect Manifold to Claude, ChatGPT or Cursor, then paste this prompt.
Compare these five creators on follower count, post cadence and engagement. Then check which hashtags they use most and pull the transcripts of their top three reels.
Where to go next
To compare other Instagram MCP servers, see Instagram MCP. To track what AI engines say about your category, pair these tools with AI visibility. For other social platforms, see Reddit MCP, TikTok MCP and LinkedIn MCP. For plans, see pricing.
