Instagram MCP: connect AI agents to Instagram data

The Model Context Protocol (MCP) is quietly becoming the standard way AI agents interact with external services. Anthropic published the spec in late 2024, and by mid-2025, over 3,000 community-built MCP servers exist on GitHub — including several for Instagram. The idea: give your AI agent structured access to Instagram data (comments, posts, analytics) so it can take actions on your behalf without brittle API wrappers.

An Instagram MCP server bridges the gap between Instagram's Graph API and any MCP-compatible AI client (Claude, Cursor, custom agents). Instead of writing API integration code yourself, you run a server that exposes Instagram capabilities as tools your AI can call.

Unlike generic AI automation posts, this guide shows real CodeWords workflows — not just theory. You'll set up an Instagram MCP server and wire it into practical automation workflows for comment management, engagement tracking, and content scheduling.

Think of MCP as a universal adapter — Instagram speaks one protocol, your AI speaks another, and the MCP server translates fluently between them.

APP: CodeWords — build AI agent workflows with 500+ integrations including social platforms, LLMs, and automation triggers.

TL;DR - MCP servers expose Instagram API capabilities as callable tools for AI agents — no custom integration code needed - Primary use cases: automated comment moderation, engagement analytics, content scheduling, and DM triage - Set up requires a Meta Developer account, Instagram Business/Creator account, and the MCP server running locally or on a cloud instance

What is MCP and why does Instagram need it?

The Model Context Protocol standardizes how AI models access external tools and data. Before MCP, every AI integration required custom code: API authentication, request formatting, response parsing, error handling. MCP abstracts this into a server-client pattern.

MCP server: Exposes "tools" (functions the AI can call) and "resources" (data the AI can read). For Instagram, tools might include get_comments, reply_to_comment, get_post_analytics, or schedule_post.

MCP client: Any AI agent that speaks the protocol — Claude Desktop, Cursor, custom agents built with the MCP SDK.

Why Instagram specifically benefits: - Instagram's Graph API is complex (multiple token types, nested pagination, rate limits) - Comment management at scale is the #1 pain point for creators and brands - AI-powered responses need real-time access to comment context - Analytics data requires aggregation that AI can summarize and act on

According to Statista (2025), Instagram has 2.4 billion monthly active users. Business accounts average 200+ comments per post on accounts with 10K+ followers. Manual comment management becomes impossible above that threshold.

How do you set up an Instagram MCP server?

Prerequisites: - Meta Developer account with an approved app - Instagram Business or Creator account - Long-lived access token with relevant permissions - Node.js 18+ or Python 3.10+ runtime

Step 1: Install the MCP server

Several community options exist on GitHub. The most maintained as of 2025: - ig-mcp — TypeScript-based, supports comments, posts, and basic analytics - instagram-mcp-server — Python-based, broader feature set including stories and DMs

Clone and install dependencies. Configure your .env with: - INSTAGRAM_ACCESS_TOKEN — Long-lived token from Meta's Graph API Explorer - INSTAGRAM_BUSINESS_ACCOUNT_ID — Your Instagram Business account ID - MCP_PORT — Port for the server to listen on

Step 2: Configure permissions

Your Meta app needs these Graph API permissions: - instagram_basic — Read profile info and media - instagram_manage_comments — Read and respond to comments - instagram_manage_insights — Access analytics - instagram_content_publish — Schedule and publish posts - pages_show_list — Required for Business accounts

Step 3: Run and verify

Start the server and verify tools are exposed correctly. Test with a simple query like get_recent_posts to confirm authentication works.

Step 4: Connect to your AI client

In Claude Desktop or Cursor, add the MCP server to your configuration. The AI now has access to Instagram tools and can call them in response to your prompts.

For integration into CodeWords workflows, the MCP server acts as a tool source within your automation pipeline — available alongside other integrations like Slack, Google Sheets, and email.

What are the practical automation use cases?

Four high-value workflows once your Instagram MCP is running:

1. AI comment moderation

The workflow: - Trigger: new comments arrive (polled every 5 minutes) - AI classifies each comment: genuine engagement, question, spam, negative sentiment, or promotional - Spam → hide automatically - Questions → generate a response draft, route to approval queue - Negative sentiment → flag for human review with suggested response - Genuine engagement → auto-like, optionally reply with a thank you variant

This handles 80% of comment volume without human intervention. The 20% that needs judgment gets surfaced with context and suggestions.

2. Engagement analytics and reporting

The workflow: - Runs daily on a schedule - Pulls metrics for all posts from the last 7 days - Calculates: engagement rate, comment sentiment distribution, best-performing content type, follower growth rate - Compares against previous week's benchmarks - Generates a narrative summary (via LLM) and pushes to Slack or email

No dashboard logins. No manual data exports. Your weekly Instagram report arrives automatically with insights, not just numbers.

3. Content scheduling with AI captions

The workflow: - Input: image/video files + brief topic notes - AI generates 3 caption variants matching your brand voice - Hashtag research via search APIs for current trending tags in your niche - Schedule post via Instagram's Content Publishing API - Track performance post-publish and feed results back for future caption optimization

4. DM triage and auto-response

The workflow: - Categorize incoming DMs: collaboration requests, customer questions, spam, personal messages - Spam → ignore/archive - Customer questions → auto-respond with FAQ answers, escalate complex queries - Collaboration requests → extract key details (follower count, niche, proposal) and push to a decision queue - Personal → notify you directly

Build these in CodeWords using scheduled and webhook-triggered workflows. The templates library has social media automation patterns you can adapt.

What are the limitations and compliance considerations?

API rate limits: Instagram's Graph API allows 200 calls per hour per user token. For most automation use cases, this is sufficient. Batch your requests and cache responses to stay within limits.

Content publishing limits: Instagram limits API-published posts to 25 per 24-hour period. Stories and Reels have separate (lower) limits. Plan your scheduling cadence accordingly.

Comment automation disclosure: Instagram's terms don't explicitly require disclosure for AI-assisted replies, but best practice is to maintain authenticity. Use AI for drafting and routing, not for impersonating personal interaction at scale.

Token expiration: Long-lived tokens last 60 days. Build token refresh into your workflow (the Meta API provides a refresh endpoint). CodeWords handles token management through its OAuth integration layer.

Data privacy: Comment data and DMs may contain personal information. If you're processing this data through external LLMs, ensure compliance with your local privacy regulations (GDPR, CCPA). Consider using locally-hosted models for sensitive data processing.

Frequently asked questions

Do I need a Business or Creator account for MCP? Yes. Instagram's Graph API (which MCP servers wrap) is only available to Business and Creator accounts connected to a Facebook Page. Personal accounts cannot access the API.

Can the MCP server handle multiple Instagram accounts? Most implementations support one account per server instance. For multiple accounts, run multiple server instances with different tokens, or use a multi-tenant MCP server that routes by account ID.

Is MCP better than direct API integration? For AI-powered workflows, yes. MCP gives your AI agent tool-calling semantics — it can decide when and how to use Instagram capabilities based on context. Direct API integration requires you to pre-define every interaction path. The pricing is the same either way in CodeWords.

What happens if Instagram changes their API? You update the MCP server. Since the MCP layer abstracts the API, your AI agent's prompts and workflows don't change — only the server's internal implementation needs updating. This is the separation-of-concerns benefit of the MCP architecture.

What MCP means for the future of social automation

MCP servers aren't just API wrappers with extra steps. They represent a shift from "write integration code for every platform" to "give AI agents access to platforms and let them orchestrate." The Instagram MCP server you set up today becomes one tool in a multi-platform agent that coordinates across Instagram, Twitter, LinkedIn, and email — all through natural language instructions.

Start with comment moderation (highest ROI, lowest risk), prove the pattern, then extend to publishing and analytics. Set up your Instagram MCP connection and build your first moderation workflow in CodeWords.