Ninjō is infrastructure for AI sales agents on Instagram, WhatsApp & every channel where you sell. Create, test, analyze and improve agents by talking to Claude, ChatGPT, Claude Code or Codex via MCP — backed by templates validated across 150+ production agents that closed real revenue ($750K generated for clients). Versioned changes with instant rollback, synthetic-conversation testing, follow-ups, keyword triggers and a built-in CRM. Zero to live in minutes. Start free with 1,000 messages.
Hey Product Hunt 👋 I'm Lolo, one of the key members at Ninjō.
Two years ago we were a tiny agency building AI sales agents by hand for creators and coaches in LatAm. Fast-forward: we run 45+ clients and 150+ agents in production, handling millions of DMs on Instagram, WhatsApp and more channels.
The only reason that's possible is that we rebuilt our entire operation on top of Claude Code. We call it Cortex: every prompt template, KPI rubric, anti-pattern and playbook we learned from real conversations lives in one system that creates, tests, analyzes and improves agents.
When MCP came out, it clicked: the product was never the dashboard. It's the infrastructure plus the accumulated intelligence. So we exposed all of it through an MCP server. Today you can open Claude, Claude Code, Codex or ChatGPT and say "build me an agent for my client's launch — qualify fast and send the payment link" — and it ships, connected to real DMs, verified, reversible.
Some real numbers from production:
• $750K+ in sales generated from our agents - 1.9M conversations - millions of messages handled
• One agent did $65K in a single 4-day launch — 839 conversations, recovering 47 declined payments one by one
• 202 sales calls booked in one month on a single mentor's Instagram
There's also Ninjō Studio (web panel: real-time conversations, built-in CRM, funnel analytics) for when you want eyes on everything — but the day-to-day runs from a chat.
For the PH community: 1,000 free messages, no credit card, with code PRODUCTHUNT.
I'll be here all day — ask me anything about running 150+ agents with 3-4 people, MCP design, or what actually converts in DMs. 🥷
Report
Hi Lorenzo,
this sounds extremely interesting and useful (and something that could help me cover the part where I, as someone who spent the last 20+ years writing software, suck) - is it safe to use this kind of product in the age where most of the content online is AI-generated and people are very quickly getting sick of seeing the usual AI patterns anywhere?
I am not trying to be annoying, I'd genuinely like to use this kind of product.
@foobar_beer Not annoying at all, it's the right question. The AI-slop fatigue is real, but I think it's mostly a content problem... mass-produced posts nobody asked for. We work 1:1 in DMs, where the person already raised their hand. And sounding human has been our whole focus from day one. We started with creators, where the agent had to represent an actual person. So we had no choice but to solve it. We've put a lot of development into exactly that, and at this point all our agents read as human.
You can build a good agent in a few minutes using the MCP! Give it a try: https://www.ninjo.ai/
Report
This looks awesome... what's the one use case that's driven the most revenue for a single client so far? Would love a real example
1. Launches. An agency client runs product launches over WhatsApp. In one 4-day launch a single agent: qualifying, handling objections, closing.
The part I didn't expect was where a real chunk came from: the agent chased declined payments one by one in the chat and recovered 47 of them. At launch peak every unanswered minute is money on the floor, and a human team physically cannot keep up with that volume inside that window.
2. Ads → DM → high-ticket call (the most repeatable one). Way more boring, and it's what most of our clients actually run: paid ads straight into IG/WhatsApp DMs, the agent qualifies and books the call, a human closes. One client sits at 200+ booked calls a month, every month, ~$200K generated. Less spectacular than a launch spike, but it compounds instead of ending on day 4.
3. The one I'm most excited about: an insurance broker. The agent quotes car insurance inside the chat, real quotes from 6 carriers, 29 options across 5 tiers, pulled live through a custom tool. You send a license plate and a postal code and get actual prices in the conversation instead of a link to a form.
Happy to go deeper on the mechanics of any of them.
Report
More than an AI setter, a whole AI comercial solution. Amazing work! Lets see the automated CRM running
Ninjō AI
Hi Lorenzo,
this sounds extremely interesting and useful (and something that could help me cover the part where I, as someone who spent the last 20+ years writing software, suck) - is it safe to use this kind of product in the age where most of the content online is AI-generated and people are very quickly getting sick of seeing the usual AI patterns anywhere?
I am not trying to be annoying, I'd genuinely like to use this kind of product.
Ninjō AI
@foobar_beer Not annoying at all, it's the right question. The AI-slop fatigue is real, but I think it's mostly a content problem... mass-produced posts nobody asked for. We work 1:1 in DMs, where the person already raised their hand. And sounding human has been our whole focus from day one. We started with creators, where the agent had to represent an actual person. So we had no choice but to solve it. We've put a lot of development into exactly that, and at this point all our agents read as human.
You can build a good agent in a few minutes using the MCP! Give it a try: https://www.ninjo.ai/
This looks awesome... what's the one use case that's driven the most revenue for a single client so far? Would love a real example
Ninjō AI
@agustin_oroquieta Thanks! Few examples here:
1. Launches. An agency client runs product launches over WhatsApp. In one 4-day launch a single agent: qualifying, handling objections, closing.
The part I didn't expect was where a real chunk came from: the agent chased declined payments one by one in the chat and recovered 47 of them. At launch peak every unanswered minute is money on the floor, and a human team physically cannot keep up with that volume inside that window.
2. Ads → DM → high-ticket call (the most repeatable one). Way more boring, and it's what most of our clients actually run: paid ads straight into IG/WhatsApp DMs, the agent qualifies and books the call, a human closes. One client sits at 200+ booked calls a month, every month, ~$200K generated. Less spectacular than a launch spike, but it compounds instead of ending on day 4.
3. The one I'm most excited about: an insurance broker. The agent quotes car insurance inside the chat, real quotes from 6 carriers, 29 options across 5 tiers, pulled live through a custom tool. You send a license plate and a postal code and get actual prices in the conversation instead of a link to a form.
Happy to go deeper on the mechanics of any of them.
More than an AI setter, a whole AI comercial solution. Amazing work! Lets see the automated CRM running
Ninjō AI
@benjamin_sonne yes! we're launching soon the agentic CRM and the new version of the UI.
Love the proposal — I'll run a test on my end and get back to you. It's a great fit for my current use case.