[00:02] and it fixes the most frustrating thing about working with AI agents. You know the moment, you ask your agent to build something, it writes a plan, you skim it, you say yes, and an hour later it has built the wrong thing. Planotator [00:16] stops that from ever happening again. Before your agent builds anything, its plan opens in your browser like a document. You highlight the bits you don't like, you leave a comment like a teacher marking homework. You press one [00:29] button and your agent reads every note and fixes the plan before it writes a single line. And it doesn't stop at plans. You can mark up the pages your agent designs, the code it changes, and the questions it has for you, all in one [00:43] place. It works with Claude Code, Codex, Gemini, Hermes, Grok Bot, and more. It's free, it's open source, and it runs on your own computer. By the end of this video, you'll know how to set up Planotator in one line, how to review [00:58] your agent's plans and pages like a pro, and how to run a whole team of agents from one inbox. And stick with me because later I'll show you the one setting that lets your agents keep working while they wait for your answer. [01:12] Let's get into it. Think of your AI agents as an orchestra. Claude Code is on the violin, Codex is on the drums, Gemini is on the keys, Hermes is on the horn, Grok Bot is on the cello. Every one of them is [01:27] talented, but an orchestra with no conductor sounds like noise. Each player goes off and does their own thing and nobody hears the wrong note until the song is ruined. Planotator turns you into the conductor. You stand at the [01:41] front, you read the score before anyone plays, and the second one of them goes off key, you point at it and they fix it. That's the whole idea. So let me show you how it actually works. Let me start with what Planotator is. [01:55] Planotator is a review screen for your AI agents. It opens in web browser, but nothing leaves your computer. Your plans, your notes, and your code stay on your machine. It plugs straight into the agent you already use through plugins [02:11] and hooks. A hook is just a little rule that says, when the agent does this, open Planform. So when your agent writes a plan, Planform opens by itself. When your agent finishes a piece of work, you can open it there too. You don't copy [02:26] and paste anything. You don't switch between 10 tabs. It all comes to you. Here's a real example. I asked Claude to plan a client onboarding portal for a software founder. New clients get a welcome checklist. They upload their [02:41] brand files. They book a kickoff call, and they can see their progress. Claude wrote a seven-step plan. And here it is inside Planform. On the left, you get a contents list so you can jump to any part of the plan. In the middle, you [02:55] get the plan itself laid out like a clean document. On the right, you get your notes. Now look at step two. Claude wants to build the whole login system from scratch. Passwords, security, password reset emails, all of it written [03:11] by hand. That's a week of work, and it's the riskiest part of the whole app. So I highlight that line. A little toolbar pops up. I click comment and I type, don't build login from scratch. Use Supabase Auth or Clerk. It's safer and [03:27] saves a week. The line turns orange and my note lands in the panel on the right. Then I scroll down to the file uploads and I give the virus scanning step a thumbs up so Claude knows that part is good. And now the button at the top says [03:41] send feedback. One click and every note goes back to Claude as its next message. Claude reads the notes, rewrites the plan, and it opens again so I can check the new version. That took me about 30 seconds. Without it, I would have found [03:56] the build. Here's why this matters so much. The plan is the easiest place to fix a mistake. Fixing one sentence in a plan takes seconds. Fixing the same mistake after the agent has built a whole app [04:11] takes hours. And when you're running more than one agent, those hours stack Now, the old way of doing this is painful. Your agent prints its plan as a wall of text in the terminal. You squint at it. You try to describe what you want [04:27] changed in one long message. Okay, in step two, the bit about passwords, don't do that. And in step four, the thing about files, that's fine, but change the part about And half the time, the agent fixes one thing and forgets the other [04:42] two. The new way is the way you already mark up a document. Highlight the line, leave the note right on it, send. The agent gets every note pinned to the gets lost. If you want to set this up properly with [04:57] the hooks, the prompts, and the review routine already done for you, that's what we do every day inside the AI Profit Boardroom. Inside, you get the Planetator agent review system I built for this video. That's the setup for [05:10] for this video. That's the setup for Claude Code, Codex, Hermes, and Grokbot. The exact review checklist I run on every plan before I let an agent build and the prompts that make your agents write plans that are easy to review in [05:22] the first place. You also get the code review routine and the inbox setup I'll show you in a minute, so your whole team of agents reports to one place. And you get the full agent OS, which connects Claude, Codex, Hermes, and Grokbot to [05:37] one shared memory, so the plans they write already know your business. There are four coaching calls every week where you can bring your exact setup and get help with it. There are daily tutorials, so when Planetator or Claude Code ships [05:50] something new, you see how to use it that week. And there's a 30-day roadmap that takes you from your first agent to a team of agents you can actually trust. Over 3000 business owners are already inside building with these tools every [06:05] description, or go to aiprofitboardroom.com. Now let's give every player in the orchestra a part. I'm going to walk you through four businesses and the exact way each one would use Annotator. [06:20] The first one is the software founder I just showed you. If you build software or you're building your first app with Claude Code, plan review is your biggest win. Here's how you turn it on. In Claude Code, you install the Annotator [06:33] plugin. That's two commands and I'll put them on screen. Then you restart Claude Code. From then on, anytime Claude goes into plan mode and writes a plan, Annotator opens in your browser automatically. You don't type anything [06:47] to trigger it. If you use Codex instead, it's even easier. The installer turns plan review on for Codex by itself. And if you use Gemini, Copilot, Open Code, or a few others, there's a setup for each one in the same installer. So [07:02] whatever agent you build with, the plan comes to you before the work starts. The second business is a marketing agency. Agencies build a lot of landing pages for clients, and AI agents are now really good at building them, but [07:17] they're not perfect. The headline is too long. The button says the wrong thing. The reviews are buried at the bottom where nobody scrolls. Normally, you'd take a screenshot, draw arrows on it, and paste it back to your agent with a [07:30] paragraph explaining what you mean. With Annotator, you open the actual page. Here, I've got a landing page my agent built for a client's ad campaign. I run one command, Annotator Annotate with the page file, and it opens the real page, [07:46] working exactly how a visitor would see it. Now I click on the headline and say, shorter, lead with the result, not the feature. I click the button and say, change this to book your free audit. I click the reviews and say, move these up [08:00] straight under the hero. Three notes pinned to the exact parts of the page. Send. The agent gets all three with the exact spot each one belongs to and fixes the page in one go. For an agency, this is the difference between five rounds of [08:15] back and forth and one. And you can do the same thing with any web page, not just ones your agent built. Give Planetator a link and it opens that page so you can market up and send the notes to your agent. So if a client sends you [08:28] a competitor's page and says make ours more like this, you can mark up exactly your agent. The third business is an online store. Say you run a store and your agent has just changed the product page. New [08:43] shipping banner, new size guide, a fix to the checkout button. Before that goes live, you want to see exactly what changed. That's what code review is for. You run Planetator review and it shows you every change side by side. Old [08:58] version on one side, new version on the other. Red for what got removed, green for what got added. You don't need to be a coder to read this. You're looking for anything that surprises you. Here I can see the agent changed the shipping [09:12] threshold text, but it also deleted a line I didn't ask it to touch. So I click that line and leave a comment. Why did you remove this? Put it back. And if you're not sure what a change does, there's an ask AI button built in. You [09:26] highlight the part you don't understand and you ask it in plain words. What does this change do to my checkout? It explains it right there. You can even launch an AI review that goes through the whole change and leaves its own [09:39] comments on the lines it's worried about. Then you send your comments back to the agent and it fixes them. It also works on GitHub pull requests. So if a developer or an agent opens a pull request on your store's code, you paste [09:52] the link into Planetator and review it the same way. The fourth business is a consultant. Consultants live in documents, proposals, client reports, strategy decks written as notes. More and more of [10:06] those first drafts are written by AI agents now. With Planatator, you can point it at any document or a whole folder of them. Here, my agent wrote a proposal for a new client. I open it in Planatator and I mark it up just like [10:21] the plan. The opening paragraph is all about the consultant instead of the client. So I highlight it and leave a note. This is all about us. Rewrite it around their problem. The scope section is too vague. So I [10:34] highlight it and say list exactly what we deliver in the first month. And the timeline is far too long. So I cross it out with the red line tool. Send and the agent rewrites the proposal with every note in place. There's even a command to [10:49] mark up the agent's last message. So if your agent just replied with a long answer in chat and you want to push back on two sentences, you open that reply in Planatator, highlight the two sentences and send your notes back. No more [11:02] retyping half of what it said just to point at it. Now here's where it goes from a handy tool to something much bigger. Everything I've shown you so far is one agent, one review. But most people using [11:15] AI seriously now have more than one agent running. Claude Code building something, Codex fixing bugs, Hermes doing admin, Grokbot handling messages. And every one of them at some point needs you, a question, a decision. [11:30] Should I use option A or option B? I've finished. Can you check this? And those questions are scattered across five different windows. So you miss one and that agent sits there doing nothing for 3 hours waiting for an answer you didn't [11:44] know it was asking. That's what the Planatator inbox fixes. The inbox is one window where all your agents leave you their messages, their questions, their files, and their finished work. You type Planatator inbox [11:59] and it opens. On the left you see every conversation grouped by what each one Questions at the top, things to review next, finished work below that. with a click. You open a file and you mark it up the [12:15] same way I've shown you all video. And when you press send, your reply goes back to the exact agent session that asked, not a new chat, the same session, so it carries on right where it stopped. And here's that setting I told you about [12:30] at the start. When an agent sends something to the inbox, nothing holds it open. It doesn't freeze while it waits for you. It sends its question and keeps working on everything else it can do or it finishes its turn. So you can answer [12:44] when it suits you. You could be on a client call, come back an hour later, clear five questions in two minutes, and every agent picks straight back up. That's the conductor's score from the orchestra. Every player's part on one [12:58] page in front of you. Claude Code, Pi, and Open Code connect to the inbox through the Planetator plugins. Hermes, Open Claw, and others connect through their own Planetator plugins too. And any other agent can [13:13] connect to it with one command because the inbox speaks the same standard language a lot of AI tools use to talk to each other. So whatever mix of agents place. Let me show you how to set the whole [13:27] thing up start to finish. Step one, open your terminal app. That's the text window on your computer where you type commands. Paste in the one-line installer from the Planetator site. I'll put it on screen and it's in the [13:40] description too. It installs Planetator, looks at your computer, finds the agents you already have, and sets each one up. Step two, if you use Claude Code, add commands from the Planetator page and restart Claude Code. Step three, ask [13:57] your agent to plan something real, not a test, a real job from your business this week. When the plan opens in your browser, highlight one thing you change and send it back. Step four, type Planetator inbox and leave that window [14:12] open while you work. That's it. Four steps and you're conducting instead of Now, you might be sitting there with a few doubts. The first one is, I'm not a few doubts. The first one is, I'm not a coder. Is this really for me? Yes. The [14:26] plan review and the document markup are just reading and highlighting. If you can leave a comment on a shared document, you can use Plannetator. And the code review has the ask AI button built in so you can ask what any change [14:39] means in plain words before you approve it. The second one is, won't this slow my agents down? It's the opposite. The review takes seconds and it saves the hours your agent would lose building the wrong thing and then undoing it. And [14:53] with the inbox, your agents don't sit frozen while they wait for you. They keep working and they pick up your answer when it arrives. The third one answer when it arrives. The third one is, is my stuff private? Plannetator [15:05] runs on your own computer. Your plans, your notes, your drafts, and your history stay local by default. It doesn't collect usage tracking. The only times anything leaves your computer are the ones you choose, like asking AI a [15:19] question about your code or opening a pull request from GitHub. So here's what you can do this week. Step one, install Plannetator with the one-line installer and let it find your agents. Step two, next time your agent [15:33] plans anything, review it in Plannetator before you say yes. Leave at least one note, even if it's a thumbs up on the part you like. Step three, when your agent builds a page or a document for you, open it in Plannetator and mark it [15:48] up there instead of writing a long message. Step four, if you run more than one agent, open the inbox and let them all report to one window, one plan at a time. And if you want help doing that, join the AI Profit Boardroom. You get [16:03] the Plannetator agent review system from this video, the setup for Claude Code, Codex, Hermes, and Grokbot, the plan review checklist, the code review routine, and the inbox setup, all with the exact prompts I use, plus the full [16:18] agent OS, so every agent in your orchestra shares one memory of your business and the plans they bring you are good before you even open them. You get four coaching calls every week, daily tutorials, and a 30-day roadmap to [16:31] go from one agent to a whole team you can trust. And you get a room of over 3,000 business owners doing exactly this. So when you get stuck, someone has already solved it. The link is in the comments and description or go to [16:46] aiprofitboardroom.com. So that's Planetator. It's a free open-source review screen for your AI agents. Your agent's plan opens in your browser before it builds anything. You highlight, comment, and send, and the [17:01] agent fixes the plan first. You can mark up pages, documents, and your agent's replies the same way. You can review every code change side by side with AI explaining anything you don't understand. And the inbox puts every [17:15] agent's questions and finished work in one window while they keep working. It works with Claude Code, Codex, Gemini, Hermes, Grokbot, and more. And it all runs on your own computer. Drop a comment and tell me, would you use [17:29] Planetator mostly to review plans before your agent builds or to run a whole team of agents from the inbox? I'll see you in the next one.