How to Use OpenClaw for Cold Outreach in 2026

What's up everybody, welcome back. Today I'm going to show you exactly how to use OpenClaw to do cold outreach that actually closes clients. This isn't theory — I'm going to walk you through the exact setup one of our users ran, and how just 10 targeted emails closed her first client on this system.

Before we get into the tech, I want to talk about copy for a second. Because most people who try to automate cold outreach fail — not because of the tech, but because of the message. 80 to 90 percent of cold emails start with "I, I, I, me, me, me." They're all about the sender. Nobody cares. The prospect cares about themselves, their business, their problems, their goals.

So the framework we're using here is value-first. Make it about them, not you. Lead with something specific you noticed about their business. Then offer to give them something — don't ask for anything. Reduce the perceived risk of engaging with you. That's it. That's the whole philosophy.

Now let's talk about the "3 specifics" method. Before OpenClaw writes a single email, it researches the prospect and finds three specific things about their business. Maybe they just hired a VP of sales. Maybe their website has a gap in their onboarding flow. Maybe a competitor just launched something in their space. Those three specifics go into the email opening. When a prospect reads something that only applies to them — that's not mass personalization, that's real personalization — the response rates are completely different.

Okay, let's get into the actual setup. There are four components you need. First, the API Gateway skill from ClawHub. This connects OpenClaw to Google — Gmail, Drive, Sheets, Docs — as well as Apollo, ActiveCampaign, Notion, and a bunch of other platforms. Install it once, authorize your Google account, and your agent can now read from your prospect spreadsheet and send emails from your actual Gmail.

Second, a Brave Search API key. Brave Search is what your agent uses to research prospects in real time. You give it a company name or a person's name and domain, and it goes out and finds recent news, blog posts, LinkedIn signals, anything publicly available. This is what powers the personalization. Get a free API key from Brave's developer portal — it takes about 2 minutes.

Third, the email validation tool. Here's how it works: you give it a first name and a company domain. It programmatically generates every common email pattern — first.last, f.last, firstlast, and so on. Then it runs an MX handshake verification. It literally talks to the company's mail server and asks, does this email exist? The server responds yes or no, without you actually sending anything. No bounces, no spam flags. It returns the one valid email address. This replaces Apollo, it replaces Clay, and it costs almost nothing.

Fourth, the Minimax M2.5 model. This is your email writing model. It's cheap — we're talking $64 for 44.7 million tokens over five days. That's for all the research, the validation, and the writing. Minimax follows instructions really well, which matters a lot when you're writing personalized emails with a specific tone and structure. If Minimax isn't available, Kimi 2.5 is a solid fallback. For your heartbeats and background monitoring tasks, free models work perfectly fine — no need to burn paid credits on those.

Now let me walk you through the actual workflow. Nikki — one of our users — set this up using her own Gmail, her own Google Drive, Google Sheets as her prospect CRM, and Google Docs for her email templates. That's all you need. You don't need Instantly for this level of outreach. Instantly makes sense if you're sending 100,000 emails. For precision outreach — ten to fifty highly targeted emails — OpenClaw with Gmail is all you need.

Her Google Sheet had columns for first name, company, domain, and a status field. The agent reads each row, runs Brave Search to research the company, generates the email candidates, validates the right one via MX handshake, writes the personalized email using Minimax, and then sends it from her Gmail. Everything gets logged back into the sheet — the validated email, the research notes, the email sent, the timestamp.

With this setup, ten emails closed her one new client. Not ten thousand. Ten. The ROI was immediate. She's now running variations of this framework for her own clients using the same value-first approach. The point is: precision outreach with real personalization beats mass volume every time.

One thing I want to hammer home: this only works if the emails are actually about the prospect. The agent can do the research, the agent can do the writing, but you have to instruct it to write for them, not about you. If you set this up and your emails still sound like generic outreach, the problem is in your prompt, not in the system. Make it specific, make it valuable, make it easy to say yes to.

The public OpenClaw documentation covers all of this step by step — how to install the API Gateway skill, how to configure Brave Search, how to set up the email validation, and how to prompt Minimax for cold outreach copy that actually converts. If you want us to build this with you, the link is in the description.

That's the whole framework. OpenClaw for cold outreach — no Clay, no expensive data tools, just precision targeting and value-first copy. Go build it.