ClawSight Blog
ClawSight turns what you can already see on your Mac into AI-ready context, so you can point, speak, and let OpenClaw act without scraping, copy-paste, or prompt gymnastics.
If you have ever tried to scrape Facebook for leads, you already know the frustration. You set everything up, run your automation, and then... nothing. Blocked. Rate limited. Broken. And LinkedIn? Do not even get me started. X is the same story. These platforms have made it their mission to stop you from pulling the data you actually need to run your business.
So you do what most people do. You copy and paste. You take screenshots. You switch tabs fourteen times. You manually explain to your AI tool what you are looking at, what you want, and why. It is exhausting, and it is slowing you down.
There has to be a better way. What if your AI could just... see what you see?
That is exactly what I built ClawSight to do.
ClawSight is a Mac application that works alongside OpenClaw, the AI assistant I use to run just about everything in my business.
Here is the simple version: ClawSight gives OpenClaw eyes on your screen and ears for your voice.
It shows up as a small companion that follows your cursor around the screen. When you hold down Control + Option and start talking, ClawSight captures what is on your screen at that moment, bundles it with your voice instruction, and sends everything over to OpenClaw. OpenClaw then takes action based on what you are looking at and what you said.
No scraping. No copy-paste. No context switching. Just point, speak, and let OpenClaw go to work.
I have leads on Facebook. I have leads on LinkedIn. I have leads on X. And every single one of those platforms has made it nearly impossible to scrape their data reliably. They block it, throttle it, and change their structure constantly just to break your tools. It is not a coincidence. It is a policy.
Traditional automation falls apart when the platforms fight back. I needed a different approach, one that did not depend on the platform cooperating.
Even when scraping works, the workflow is clunky. You are constantly feeding your AI tool information manually: paste this URL, upload this screenshot, switch tabs, explain the context again. Every one of those little interruptions pulls you out of focus. The copy-paste tax is real, and it adds up fast across a workday.
Here is the thing. Most of the people I work with are small business owners. They are brilliant at what they do, but they did not sign up to learn agent architecture or figure out how to wire up complex AI workflows. They just want help. They want something that feels natural: look at a thing, talk to it, get a result.
That is the gap ClawSight was built to fill.
I actually came across an open-source project called Clicky, a small cursor-following pointer that @FarzaTV built and released under the MIT license. It was cute. It followed your cursor around, you could talk to it, but it did not really do much. The developer added little agents behind it, which was cool, but I did not want those agents. I wanted my OpenClaw, because my OpenClaw is brilliant and I trust it. So I started thinking: what if I could wire this little cursor thing directly into OpenClaw? What if it could also pass along the Obsidian note I had open, or the web page I was looking at?
And just like that, ClawSight was born.
ClawSight runs quietly on your Mac in the background. It appears as a small visual companion that follows your cursor wherever you go. It is lightweight. It is always there. It does not get in your way.
When you are ready to use it, you hold down Control + Option and start speaking. That is it. The moment you do that, ClawSight takes a screenshot of what is currently on your screen, captures your voice instruction, and sends both to OpenClaw, along with the web address or Obsidian note.
You are not uploading anything manually. You are not typing a prompt. You are just looking at what you are working on and talking, the same way you would explain something to a colleague sitting next to you.
OpenClaw receives the screenshot and your instruction together. It understands the context from what is visually on your screen, and then it acts. It can respond, organize information, draft messages, update your database, send emails, create notes, whatever you have set it up to do. The screen becomes the input. Your voice becomes the command. OpenClaw does the rest.
First principles thinking: instead of trying to extract data from platforms that resist it, ClawSight uses what is already visible on your screen. If you can see it, ClawSight can work with it. That means it does not matter whether a platform blocks scraping. If you can look at a profile, ClawSight can capture it.
No more copy-paste cycles. No more tab switching. No more manually explaining what you are looking at. You stay in your work. Your AI comes to you rather than you having to drag information over to it.
You do not need to understand how OpenClaw works under the hood to use ClawSight. As the operator, I set up OpenClaw for my clients. They just use ClawSight. They hold Control + Option, they talk, and things happen. That is the whole experience from their end. Simple, approachable, and genuinely useful.
You are looking at a Facebook profile or a LinkedIn page. You hold Control + Option. You say: "Add this person to our people database and draft a personalized email based on our previous conversations." ClawSight captures the screen, sends it to OpenClaw with your instruction, and OpenClaw handles the rest. No scraping. No copy-paste. Done.
Here is an example of how I use it on LinkedIn.
Move people through your pipeline by speaking to ClawSight while you look at your Kanban board. Ask OpenClaw to trigger next actions, update stages, or draft follow-ups based on prior conversation history. Your CRM starts to feel less like admin and more like a conversation.
This one is my personal favorite. I used ClawSight to run an outreach campaign to everyone enrolled in my school. I just said: "Make me a Kanban with four channels, take all the people who are enrolled in my school and put them in the approach channel." Because my OpenClaw knows my entire company and the state of every person in my database, it just went and did it. Then I said: "Write a letter to each one of them, but take into account every conversation I have ever had with them." OpenClaw set up seventy-three draft emails for me. Seventy-three. And then it asked: "Do you want me to send these?"
I said: "Slow your roll. Let me look at them first."
Because I am a control freak. And that is completely fine, because the system supports that. I reviewed each one, made small tweaks, and sent them out through my Google account. Seventy-three semi-personalized outreach emails, done.
That is what AI-assisted work is supposed to feel like.
Use ClawSight while you have an Obsidian note open. The current screen context, including the note you are actively working in, gets passed to OpenClaw along with your voice instruction. You can build better session notes, capture insights before they disappear, and keep a record that actually reflects what you were thinking and working on.
This is the big picture goal. Business owners should not need to understand what is happening behind the scenes. As the operator, I configure OpenClaw for each client. I set up their workflows, their database, their automations. Then I give them ClawSight. They just talk to it. They do not need to know about prompts or agents or API calls. They just work, and their AI helps them.
One of the things that drove me crazy about voice-based AI interaction was losing the transcript. You have a great conversation, get real work done, and then it is just gone. No record. No reference.
So I built Telegram mirroring into ClawSight. The entire conversation, both sides of it, echoes into Telegram in real time. If you want to create a session note or just keep a record of what happened, it is all there. You never lose the voice into oblivion.
I also had to solve the continuity problem so that OpenClaw does not just reset and forget everything every time you pick up a new conversation. And then I started thinking: what if you want the Telegram session and the ClawSight session to be the same session? But what if you want them to be separate, so you can send the Telegram agent off to handle something while you keep working locally on something else?
So I built a toggle. You can mirror them together or split them apart depending on what you need. It is a small feature, but it opens up a lot of workflow possibilities.
I am also considering adding Apple Messages as a channel, because most of my clients are Mac people and not everyone wants to use Telegram. If that sounds useful to you, I would love to hear what you think.
Picture this. You are sitting at your Mac. You have a list of leads open on your screen, people you want to reach out to about a new offer you are running.
You hold Control + Option.
You say: "Add this person to our people database and draft an email for them based on our past conversations and the new offer we are running."
ClawSight takes a screenshot. It captures the person you are looking at on your screen. It bundles that with your voice instruction and sends everything to OpenClaw.
OpenClaw checks your database. It pulls up your conversation history with that person. It looks at the new offer details. It writes a personalized draft email.
It comes back to you: "Draft is ready. Want me to send it?"
You say: "Not yet, let me read it first."
You review it. You change one line. You say: "Okay, send it."
It is sent.
The whole thing took about forty-five seconds. You stayed focused. You never opened your email client, copy-pasted a single thing, or stopped what you were doing. You just pointed, spoke, and let your AI handle it.
That is the ClawSight experience. Speed, natural interaction, and you staying in control the whole time.
Here is the bigger picture. AI is genuinely powerful right now. But most interfaces for using it are still awkward. You are either writing long prompts, switching between tools, or learning workflows that feel more like programming than working.
ClawSight is a move toward something more natural. See what I see. Hear what I say. Help me act. That is the model. And when you lower the barrier enough, when you make it that simple, suddenly the people who could benefit most from AI are actually able to use it. Not just the tech-savvy early adopters. Everyone.
That matters to me. A lot.
I have only shown ClawSight to a handful of people so far. The first person I demonstrated it to looked at it for about two minutes and said: "Yes, I want to be an affiliate for this. This is freaking cool. This would solve so many problems."
That reaction told me I was onto something real.
This is not the finished product. It is the beginning of a direction. AI that works with your actual screen and your actual workflow, without asking you to become a prompt engineer. An assistant that can genuinely see what you are working on, hear what you need, and get it done.
The early tester feedback is already shaping where ClawSight goes next. And I want more of it.
Right now, ClawSight is a Mac application. If you are on macOS and you use OpenClaw, or you are curious about what OpenClaw could do for your business, ClawSight is designed to plug right in.
Early distribution is happening through direct download for testers. You download the app, install it, and start using it. There is no App Store submission required for this stage, which means you can get it fast and I can iterate quickly based on your feedback. If you are an early tester and something needs to change, we can turn that around without waiting weeks for approval cycles.
This approach lets me keep ClawSight sharp, responsive, and genuinely useful rather than locked into a slow release process while the tool drifts out of sync with how OpenClaw is evolving.
If you are nodding your head right now, you are exactly who I built this for.
Here is what I want you to take away from this. AI is not just for the technically savvy anymore. It is not just for developers or founders with dedicated ops teams. It is for the business owner who has forty-five minutes between meetings and a hundred things on their list. It is for the person who has been curious about AI but every tool they have tried has felt like a project in itself.
ClawSight is not a project. It is a shortcut. It is the thing that sits quietly on your screen, follows your cursor, and waits for you to hold down two keys and say what you need.
You do not have to figure out how it all works behind the scenes. You just have to look, speak, and let it go.
That is what AI-assisted work should feel like. And we are getting there.
If you want to see whether ClawSight fits the way you actually work, I would love to show you. No pressure, no pitch deck, just a real look at the tool and a conversation about whether it makes sense for you.
Book a short discovery call here: https://calendar.notion.so/meet/marnimelrose/30minreview
Let's find out together... and share this with anyone who you think might need it.
P.S. I am even thinking about ClawSight Cloud where we host your OpenClaw for you. What do you think about that?