About Ninjō AI
Ninjō AI is infrastructure for building and running AI sales agents on messaging channels like Instagram and WhatsApp. It operates through an MCP server that lets users create, test, and improve agents from within Claude, Claude Code, ChatGPT, or Codex. The platform built on top of Claude Code includes system they call Cortex: a repository of prompt templates, KPI rubrics, and playbooks gathered from production deployments.
Review
Ninjō launched this week, so it's a young product with a clear starting point. The tool's key idea is exposing its internal agent-running system through MCP, which lets you direct the flow from a chat instead of a dashboard. The reported numbers from production are substantial: $750K+ in sales generated, 1.9M conversations, and millions of messages handled across 150+ agents for 45+ clients.
Key Features
- MCP server integration: Build and deploy agents by talking to Claude, Claude Code, ChatGPT, or Codex via Model Context Protocol.
- Versioned changes with instant rollback: Agent configurations are versioned so you can revert at any time.
- Synthetic-conversation testing: Run simulations before pushing agents live against real DMs.
- Follow-ups and keyword triggers: Agents can initiate follow-up messages and respond to specific keywords in the conversation.
- Ninjō Studio (web panel): Real-time conversation viewing, built-in CRM, and funnel analytics for oversight, though daily operations run from a chat interface.
Pricing and Value
The free tier includes 1,000 messages with no credit card required, expanded to "14 days or 1k DM's Free" as an alternative promotion. The Product Hunt launch code "PRODUCTHUNT" activates the 1,000 free messages. Beyond the free tier, the page does not list subscription pricing, so the paid cost structure is not yet defined. The value case rests on the reported outcomes: one agent booked 202 sales calls in a month for a mentor, one client generated roughly $200K from 200+ booked calls per month, and a 4-day launch agent recovered 47 declined payments that contributed to a $65K run.
Pros
- Agent creation happens in natural language, letting you describe requirements and have a working agent shipped to real channels.
- Templates and playbooks come from 150+ production agents that closed real revenue, giving you experience baked into the setup.
- The insurance quoting example shows custom tools can pull live data into conversations, calculating real quotes from 6 carriers with 29 options across 5 tiers.
- Instant rollback on versioned changes protects against breaking modifications to agents that are handling live conversations.
Cons
- The current setup depends on Claude Code as the operational backbone, so its workflow is inseparable from Anthropic's toolchain.
- Agentic CRM and the new UI version are announced as launching soon; neither is live right now.
- Not well suited for teams who lack technical capacity and need a purely visual, dashboard-first setup, since agent creation through MCP requires familiarity with conversational programming and Claude Code.
Ninjō AI fits agencies and businesses that operate sales conversations at high volume on messaging channels and don't want to hand-build agent infrastructure. Users comfortable with Claude Code will find the MCP approach direct; those wanting a no-code editor may want to wait for the planned UI version. The production numbers demonstrate the system's capacity, but prospective users should weigh the current reliance on coding via chat against their team's administrative abilities. So that's a worthwhile option for sales teams that treat DMs as their primary channel and want an infrastructure they can direct with a chat window.
Open 'Ninjō AI' Website
Your membership also unlocks:
