Reply to Comments on Instagram
The Reply to Comments on Instagram agent automatically detects new comments on your posts and generates on-brand replies using your custom AI instructions. It can handle simple engagement, answer FAQs, or direct users to your link-in-bio.
Setup (Step-by-Step)
Connect Your Account (New Users)
Click the Connect Instagram button.
You will be taken to a Facebook/Instagram OAuth screen.
Approve all requested scopes:
instagram_business_basic,instagram_business_content_publish, andbusiness_manage_comments. This ensures all features work now and in the future.
Enable Webhooks (Existing Users)
If you've already connected Instagram for other agents, a Create Webhook button will appear.
Click it to approve the final, missing scope:
business_manage_comments. This action automatically subscribes you to the comment webhooks.
Write Your Agent Prompt
Use the editor (up to 10,000 characters) to define your agent's tone, rules, and knowledge.
Tip: We highly recommend editing the default template (see Best Practices below) by filling in your
[NICHE],[LINKTREE_OR_SITE],[BUSINESS_EMAIL], etc.
Configure Reply Settings
Top comment only: Set to 'Yes' to only reply to top-level comments. Set to 'No' to allow the agent to reply to nested replies.
Delay (in minutes): Enter 0 for an instant reply. To feel more natural, add a delay (e.g., 15) before the bot responds (max 1,000 minutes).
Test the Output
In the right-side tester, input a sample Post Caption, Username, and Comment.
Click Test Output (costs 1 credit).
Iterate on your prompt until the replies are perfect.
Go Live
Click Start Agent.
Your agent is now live and will automatically reply to all new comments on your profile.
Configuration
These are the settings located on the left side of the Agent Builder.
Instagram Profile
Select the Instagram Business account you wish to connect.
First time setup: You must click "Create Webhook" to authorize the
business_manage_commentspermission. This allows the agent to "listen" for new comments.
Top Comment Only
This setting controls how deep the agent goes into a conversation thread.
Yes: The agent will only reply to the main (root) comment on a post. It will ignore replies to replies.
No: The agent will reply to every comment in the chain.
Note: Use "No" with caution to avoid over-engaging in long debates.
Delay (in minutes)
Controls how quickly the AI posts the reply after the user comments.
0: Instant reply (usually within seconds).
1–1000: Adds a delay to make the interaction feel more human and organic.
Agent Prompt
This is the "brain" of your agent. You can write up to 10,000 characters defining your persona, tone, and rules.
Dynamic Variables: You can use these placeholders in your prompt to inject real-time data:
{commenter_handle}: The username of the person replying (e.g., @wesdaco).{comment}: The actual text of the incoming comment.{instagram_post}: The caption of the specific post they commented on. (Crucial for context-aware replies).
Prompt Tip:
"Always check the
{instagram_post}context before replying. If the post is sad, do not use happy emojis. If the user asks a question, check the FAQ section below."
Testing Your Agent
The panel on the right side allows you to simulate a comment to see how your prompt handles it before you go live.
Instagram Post: Paste a sample caption here (e.g., "Loving the fall vibes in NYC! #ootd"). This simulates the
{instagram_post}variable.Commenter Handle: Simulate a username (e.g., @fan_account).
Comment: The test message (e.g., "Where did you get that jacket?").
Clicking Test Output allows you to tweak your prompt until the tone is perfect.
Important Notes & Limitations
Self-Reply Protection
The agent is hard-coded to ignore comments made by the connected account. It will never get stuck in a loop replying to itself.
OAuth Permissions
If the agent is not firing, ensure you have granted all three required permissions:
instagram_business_basicinstagram_business_content_publishinstagram_business_manage_comments(The specific webhook permission).
Privacy
We process the comment text and username solely to generate the reply. We do not store conversation history long-term.
