ManyChat for agents
You talk. Your agent builds. The engine runs.
DMParrot is the ManyChat playbook rebuilt for agent-native authoring. Humans never open a node editor. Agents call MCP tools; a plain rule engine delivers Instagram DMs.
- 01
Start a trial + connect Instagram
Pick Starter or Growth on Pricing (card required for the 14-day trial). Sign in with Google, then link your Instagram professional account via Meta OAuth.
- 02
Point your agent at DMParrot MCP
Paste the remote into Claude Desktop, Claude Code, or ChatGPT. Your agent gets the full ManyChat-style tool surface — funnels, tags, broadcasts, analytics.
- 03
Ask for the funnel. Go live.
Describe the growth loop in chat. Your agent calls create_comment_to_dm_flow (or wires nodes itself), publishes, and matching Instagram comments get private-reply DMs. No LLM in the send path.
MCP tool surface
What ManyChat puts on a canvas, agents get as tools.
Mid-level MCP tools — expressive enough for real funnels, constrained enough that your agent can't invent something the runtime can't execute.
-
create_comment_to_dm_flowOne-shot keyword → DM → tag funnel -
create_dm_keyword_flowDM keyword auto-reply + optional tag -
create_story_reply_flowStory reply → DM growth loop -
simulate_flowDry-run before publishing -
send_broadcastSegment blast inside the 24h window -
publish_flowTake the funnel live