# Getting started You operate your Ninjo agents by **talking to Claude** β€” in plain English (or Spanish). No code, no spreadsheets, no commands to memorize. You describe what you want; Claude does it and shows you the result. This page gets you connected and shows you the rhythm. Every other page is a catalog of things you can ask for, each with a sentence you can copy and paste. --- ### Connect for the first time **What you get:** Claude linked to your Ninjo account, so it can see and manage your agents. **When to use it:** Once, the first time you open this project. After that it stays connected. **Just say:** > /mcp A browser window opens to the Ninjo login. Enter your Ninjo credentials and you're in β€” it's saved for next time. If you ever get logged out, run `/mcp` again. > **Not on the terminal?** The steps are a little different in the Claude app and on the web β€” see [Connecting](connecting.md) for the simple, step-by-step version for each. **Tips:** - You don't need an API key, a `.env` file, or any database setup. Just your Ninjo login. - To confirm it worked, try the next one πŸ‘‡ --- ### See all my agents **What you get:** A list of every agent on your account, so you know what you're working with. **When to use it:** Right after connecting, or any time you forget an agent's exact name. **Just say:** > Show me all my agents **What happens:** Claude lists your agents by name. From there you can ask about any of them by name β€” "how is *mauristacks* doing?", "test *bienestar*", and so on. --- ### Connect Instagram, WhatsApp or Facebook (and point it at an agent) **What you get:** your social accounts linked to Ninjo and assigned to the agent that should handle them, so the right agent answers on the right channel. **When to use it:** first-time setup, adding a new channel, or pointing a freshly connected account at a specific agent. **Just say:** > Show me my connected accounts, and assign my Instagram @handle to [agent name]. **What happens:** Claude lists the Instagram, WhatsApp, and Facebook accounts connected to your Ninjo and wires the one you choose to the agent you name, confirming before it saves. (Linking a brand-new account for the very first time happens in Ninjo's own setup screen β€” once it's connected, Claude can see it and assign it.) **Tips:** - One agent can own a channel, and different agents can own different channels β€” just tell Claude which goes where. - Not sure what's connected? "What accounts are connected to my Ninjo?" gives you the list. --- ### Understand the loop **What you get:** The mental model for everything else. Operating an agent is a simple cycle you repeat as often as you like: > 🌱 **Create** β†’ πŸ§ͺ **Test** β†’ πŸ“Š **Analyze** β†’ πŸ”§ **Improve** β†’ β†Ί (repeat) - **Create** β€” build a new agent from your information ([Creating agents](creating-agents.md)). - **Test** β€” run it through example chats before real people see it ([Analyzing & improving](analyzing-and-improving.md)). - **Analyze** β€” look at how it's doing with real leads ([Analyzing & improving](analyzing-and-improving.md)). - **Improve** β€” apply one focused fix based on what you saw ([Analyzing & improving](analyzing-and-improving.md)). You don't have to do all four every time. Most days you'll just analyze and improve. --- ### Ask in your own words **What you get:** Freedom from memorizing anything. There are named shortcuts (`/cortex-create`, `/cortex-analyze`, and so on), but **you never have to use them.** Just describe the outcome you want and Claude picks the right path: > *"My agent keeps sending the price too early β€” can you fix that?"* > *"How many calls did we book this week?"* > *"Add a keyword 'GUIDE' that sends my free PDF."* Every use case in this wiki gives you a **"Just say:"** line β€” that's a ready-to-paste example. Tweak the wording however feels natural; you can't break anything by phrasing it differently. **Tips:** - Claude always confirms before doing anything that changes a live agent, and it can undo a change if it made things worse. It's safe to experiment. - If you're not sure what's possible, just ask: *"What can you help me do with this agent?"* --- ### Tell Claude to report a bug or a missing feature **What you get:** if a tool fails or something you need doesn't exist yet, Claude files it with the Ninjo team for you β€” you don't have to track anyone down. **When to use it:** anytime something seems broken, or you wish Ninjo could do something it can't yet. Reporting it is how it gets fixed or built. **Just say:** > That didn't work the way I expected β€” can you report it to the Ninjo team? **What happens:** Claude writes up the issue with the details that matter and sends it to the team, which turns it into a ticket. You get a confirmation, and the team takes it from there. **Tips:** - Claude still tries to solve your problem first β€” reporting it doesn't replace getting you an answer right now. - Feature ideas count too: *"I wish I could do X β€” can you pass that along?"* --- ## Where to next | I want to… | Go to | | ---------- | ----- | | Build a brand-new agent | [Creating agents](creating-agents.md) | | Decide which words make my agent reply | [Keywords & triggers](keywords-and-triggers.md) | | See how my agent is doing and fix it | [Analyzing & improving](analyzing-and-improving.md) | | Take over a chat or hand it back | [Live conversations](conversations.md) | | Follow up and get notified automatically | [Automation & follow-ups](automation-and-follow-ups.md) | | Make images or manage what my agent shares | [Content & media](content-and-media.md) | | Track lead info and spot problems at scale | [Custom data & insights](custom-data-and-insights.md) | ← Back to the [wiki home](index.md)