What If Your Sales Assistant Never Slept?
Imagine having a team member who monitors your inbox every thirty minutes, researches every prospect before you talk to them, drafts personalized outreach based on real context, follows up automatically when conversations go cold, and prepares a briefing doc before every meeting. That is what you are going to build.
Most businesses pay $500 to $2,000 a month for commercial AI SDR tools that do some of this. The setup in this guide costs $20 to $35 a month total — $5 to $10 for a small cloud server and $15 to $25 for the AI API usage. You own everything, the data stays on your server, and the system keeps running whether you are paying a consultant or not.
This guide walks you through the entire setup from scratch, including every template and config file you need. You do not need to be a developer, but you should be comfortable following instructions in a terminal. If you can copy and paste commands, you can do this.
Download the Templates
Everything in this guide is ready to copy and paste. We have also made the key files available as direct downloads so you can drop them straight into your OpenClaw workspace:
- SOUL Sales Template (SOUL.md) — the AI personality and sales rules
- Heartbeat Template (HEARTBEAT.md) — inbox monitoring automation config
- Email Templates (outreach.md) — cold email templates, follow-ups, and response handlers
- Setup Commands Reference — every terminal command in one file
What You Will Have When You Are Done
By the end of this guide, your OpenClaw setup will handle five core jobs:
- 24/7 inbox monitoring. Every thirty minutes, the system scans your unread emails, categorizes them by urgency, and drafts responses for anything that needs immediate attention.
- Prospect research on autopilot. Before any outreach, the AI researches the company, finds recent news, and identifies talking points you can actually use in conversation.
- Personalized cold outreach. No more generic templates. Every email references something specific about the prospect — a recent hire, a project they announced, or a problem they described publicly.
- Automatic follow-ups. When someone does not reply, the system drafts a contextual follow-up based on the original conversation. Not a "just checking in" email — a real message that adds value.
- Meeting prep briefings. Before every call, you get a one-page doc with everything you need: who they are, what they care about, their recent activity, and suggested talking points.
What You Need Before Starting
Before you begin, gather these four things:
- A cloud server. Any VPS provider works — DigitalOcean, Hetzner, Linode, or whatever you already use. A basic $5 to $10 per month plan with 2GB of RAM is enough. Ubuntu or Debian is recommended.
- Basic terminal comfort. You will be running commands in a terminal window. If you have ever used SSH or installed software from the command line, you are good.
- A Gmail account. This is the email account your AI assistant will monitor and send from. A Google Workspace account works best for business use.
- An Anthropic API key. This powers the AI brain behind your assistant. Sign up at anthropic.com, add some credits, and grab your API key.
Time required: One to two hours for the full setup.
Throughout this guide, values you need to replace with your own are written in UPPER_SNAKE_CASE (like YOUR_EMAIL or YOUR_NAME). Bracket placeholders like [First Name] and [Company] are dynamic values the AI fills in per-email — you do not change those.
Part 1: Getting OpenClaw Running
Install Docker on Your Server
SSH into your cloud server. Update the system and install Docker:
- Update system: sudo apt update && sudo apt upgrade -y
- Install Docker: curl -fsSL https://get.docker.com | sh
- Add yourself to Docker group: sudo usermod -aG docker $USER
- Log out and back in for the group change to take effect
Verify Docker is running: docker --version
Install OpenClaw
Pull the image and start the container:
- Pull: docker pull openclaw/openclaw:latest
- Run: docker run -d --name openclaw --restart unless-stopped -v ~/.openclaw:/data/.openclaw -p 18789:18789 -e ANTHROPIC_API_KEY=YOUR_API_KEY openclaw/openclaw:latest
- Check logs: docker logs openclaw
Get your gateway token: docker exec openclaw cat /data/.openclaw/.env | grep OPENCLAW_GATEWAY_TOKEN
Open your browser to http://YOUR_SERVER_IP:18789 and enter the token. Then run the onboarding wizard: docker exec -it openclaw openclaw onboard
That is it for the basic install. You now have OpenClaw running on your own server.
Lock Down Your Server
Before doing anything else, secure your setup.
Install Tailscale for private networking. Tailscale creates an encrypted tunnel between your devices and your server, so only you can access OpenClaw. Install it with: curl -fsSL https://tailscale.com/install.sh | sh then sudo tailscale up. Note your Tailscale IP with tailscale ip -4.
Configure OpenClaw to only listen via Tailscale:
- docker exec -it openclaw openclaw config set gateway.bind "loopback"
- docker exec -it openclaw openclaw config set gateway.tailscale.mode "serve"
- docker restart openclaw
Set up a firewall using UFW:
- sudo apt install -y ufw
- sudo ufw default deny incoming
- sudo ufw default allow outgoing
- sudo ufw allow in on tailscale0 to any port 22
- sudo ufw enable
Install Fail2ban to block brute-force SSH attempts: sudo apt install -y fail2ban and enable it with sudo systemctl enable --now fail2ban
These three steps — Tailscale, firewall, and Fail2ban — take about fifteen minutes total and make your setup significantly more secure than most commercial tools.
Part 2: Connecting Your Gmail
This is the most important integration. Your AI assistant needs to read and send email on your behalf.
Set Up Google Cloud OAuth
- Go to console.cloud.google.com and create a new project called "SalesAI"
- Enable these four APIs: Gmail API, Google Calendar API, Google People API, Google Drive API
- Configure the OAuth consent screen — choose "External", add your email, and add the Gmail, Calendar, and Contacts scopes
- Add yourself as a test user
- Create OAuth credentials — choose "Desktop app", then download the JSON file
Connect Gmail to OpenClaw
Copy your credentials file to the container: docker cp ~/Downloads/client_secret.json openclaw:/data/.openclaw/workspace/
Then inside the container:
- gog auth credentials client_secret.json — register the credentials
- gog auth add YOUR_EMAIL@GMAIL.COM --services gmail,calendar,contacts,drive --remote --step 1 — starts the auth flow
- Copy the auth URL, open it in your browser, approve access, then copy the redirect URL
- gog auth add YOUR_EMAIL@GMAIL.COM --services gmail,calendar,contacts,drive --remote --step 2 --auth-url "PASTE_REDIRECT_URL" — completes the flow
Set the keyring password for headless operation: export GOG_KEYRING_PASSWORD="your-secure-password"
Test it: gog gmail search "newer_than:1d" --max 5 --account YOUR_EMAIL@GMAIL.COM — you should see your recent emails.
Token Expiration Warning
If your Google Cloud project is in "testing" mode (the default), tokens expire every seven days. When this happens, gog commands will fail with "Token has been expired or revoked." Just re-run the two-step auth flow above. Set a weekly calendar reminder, or submit your app for Google verification for permanent tokens.
Part 3: The SOUL File — Your AI's Sales Personality
This is the most important configuration file. The SOUL file tells your AI who it is, how to behave, and what to never do. Download the full template here.
Save this as /data/.openclaw/workspace/SOUL.md in your OpenClaw container.
What Is in the SOUL File
Core identity — defines the AI as your SDR:
- Monitor inbox for new leads and hot prospects
- Research companies and contacts before outreach
- Draft personalized cold emails and follow-ups
- Prep for calls with context and talking points
- Track conversations and remind about follow-ups
- Learn from your feedback over time
AI personality traits:
- Direct and to-the-point — sales people hate fluff
- Optimistic but realistic
- Data-driven — shows numbers, trends, patterns
- Proactive — suggests actions without waiting to be asked
- Coachable — remembers corrections
Sales philosophy and outreach rules:
- Personalization over volume. One great email beats 100 generic ones.
- Research the prospect before writing. What do they care about?
- Keep it short. Busy people skim.
- Clear CTA. Every email ends with one specific ask.
- Follow up consistently. Most deals happen after the third-plus touch.
Tone for outreach:
- Conversational, not corporate
- Curious, not pushy
- Helpful, not salesy
- Assume they are smart and busy
Email structure (every email follows this):
- Hook — why am I emailing you specifically?
- Value — what is in it for them?
- Ask — clear, low-friction next step
What the AI must never do:
- Never send an email without showing you a draft first
- Never pitch features — pitch outcomes
- Never use buzzwords like "leverage," "synergy," or "innovative solution"
- Never promise things you cannot deliver
- Never guess pricing — ask you first
Output formatting — the SOUL file also defines exactly how the AI formats its output:
- Email drafts include the TO, SUBJECT, body, and a CONTEXT section explaining the research and reasoning
- Inbox summaries categorize as HOT (reply today), WARM (reply this week), and COLD (ignore)
- Call prep docs include who they are, what they care about, talking points, expected objections, and the ask
The USER File
You also need a USER.md file with your specific details. Save this to /data/.openclaw/workspace/USER.md and fill in your information:
- Your name, role, and timezone
- What you sell and to whom
- Your ideal customer profile
- Typical deal size and sales cycle
- Your email tone preference and signature
- Your calendar link for scheduling
- Current priorities and focus areas
The more specific you are, the better the AI performs. Instead of "we help businesses grow," write "we build automation systems for B2B companies doing $2M to $50M in revenue that are still prospecting manually."
Enable Web Search
Get a free Brave Search API key at brave.com/search/api, then configure it:
- docker exec openclaw openclaw config set tools.web.search.apiKey "YOUR_BRAVE_API_KEY"
- docker restart openclaw
Part 4: Inbox Autopilot — The HEARTBEAT File
The HEARTBEAT file tells your AI to check your inbox on a schedule and categorize everything. Download the full template here.
Save this as /data/.openclaw/workspace/HEARTBEAT.md and replace YOUR_EMAIL@DOMAIN.COM with your actual email.
How It Works
Every thirty minutes, the system runs this sequence:
Step 1: Scan — searches for unread emails from the last two hours (up to 20 messages)
Step 2: Categorize each email:
- HOT — inbound leads, replies to outreach, meeting requests, pricing questions, referral introductions. These need immediate action.
- WARM — questions, interested-but-not-urgent messages, newsletters from competitors, industry news. These go into the daily summary.
- COLD — spam, marketing blasts, newsletters, anything irrelevant. These get ignored.
Step 3: Act on HOT emails — the AI reads the full email, researches the sender and their company, drafts a reply for your review, and pings you immediately.
Step 4: Summarize WARM emails — notes what they need and suggests a reply or next action.
Daily Summary
Every morning at 8 AM, the AI compiles:
- Inbox status — how many unread, how many hot leads
- Follow-ups due today
- Any deals that have not moved in 5+ days
- Your meetings for the day with prep notes
Enable the Heartbeat
Run these commands to activate 30-minute monitoring:
- docker exec openclaw openclaw config set agents.defaults.heartbeat.every "30m"
- docker restart openclaw
Test it manually: docker exec openclaw openclaw agent --message "Read HEARTBEAT.md and run the inbox check now" --deliver
Part 5: Cold Email Templates
These are the actual templates your AI uses for outreach. Download all templates here.
Save the file to /data/.openclaw/workspace/templates/outreach.md in your container.
Template 1: Problem-First Approach
Use when: you know a specific pain point they have.
Subject: [Pain point] at [Company]?
Body: Hi [First Name], I noticed [specific observation about their company or recent news]. Quick question — how are you currently handling [pain point]? Most [role/industry] teams we work with struggle with [specific challenge]. We built [brief value prop] that helps with exactly that. Worth a 15-minute call to see if it is relevant for [Company]? — YOUR_NAME
Template 2: Social Proof
Use when: you have a customer similar to their company.
Subject: How [Similar Company] solved [problem]
Body: [First Name], [Similar Company] was dealing with [pain point] — sound familiar? We helped them [specific outcome / metric] in [timeframe]. I am not saying we can do exactly the same for [Their Company], but if [pain point] is on your radar, might be worth a quick chat? 15 minutes this week? — YOUR_NAME
Template 3: Straight Pitch
Use when: you are highly confident they need this.
Subject: Quick intro
Body: Hi [First Name], YOUR_COMPANY helps [ICP] with [problem]. We work with [3-4 similar companies]. Typical results: [outcome]. If [pain point] is a priority, let us talk. Calendar link: [link] — YOUR_NAME
The Follow-Up Sequence
- Day 3 — The Bump: "[First Name] — just bumping this up your inbox in case you missed it. Still curious if [pain point] is something [Company] is thinking about."
- Day 5 — Last Try: "I will stop bugging you after this, promise. If [problem] is not a priority right now, totally understand. But if it is and I can help, let us talk: [calendar link]"
- Day 7 — The Breakup: "Have not heard back, so I am assuming now is not the right time. I will check back in [3/6/12 months] unless you tell me otherwise. P.S. — If anything changes, my email is open."
Response Handling
The templates also include responses for three common scenarios:
- "What does it cost?" — Do not answer with a price. Steer toward a call. "Great question. Pricing depends on [variable factors]. A quick call to understand your setup would let me give you an accurate number. 15 min to chat?"
- "Send me more info." — Do not send a generic PDF. "Absolutely. Before I flood your inbox, what specifically are you most curious about? [Option 1], [Option 2], [Option 3]? Or happy to just hop on a quick call."
- "Not right now." — Find out when. "Totally fair. When should I check back in? And is there anything specific that would need to change for this to become a priority?"
Personalization Checklist
Before any outreach goes out, the AI checks these five things:
- Does the email reference something specific about the prospect's company?
- Is the pain point tailored to their situation, not generic?
- Is it under 100 words?
- Is there exactly one clear call to action?
- Are the names and company details spelled correctly?
If any are missing, the draft gets revised before you see it.
Part 6: Meeting Prep on Autopilot
Every meeting should feel like you have done your homework. The AI makes sure you always have.
How Meeting Prep Works
When you have a call on your calendar, the system automatically:
- Pulls the attendee information from your calendar event
- Searches your email history for any prior conversations with that person or company
- Researches the company online — recent news, funding announcements, leadership changes, job postings
- Generates a one-page briefing doc with everything you need
What Is in the Briefing Doc
- Who they are: title, background, LinkedIn summary, how long they have been at the company
- What they care about: recent company initiatives, news, challenges, industry trends
- Conversation history: summary of past emails, calls, and previous interactions
- Talking points: three to five specific things to bring up based on research
- Objections to expect: likely pushback and how to handle it
- Questions to ask: discovery questions relevant to their situation
- Close / next steps: what you want to get out of the call and the ideal next action
Setting Up Automatic Prep
Add this to the end of your HEARTBEAT.md file:
Meeting Prep (Daily at 6 PM): Check tomorrow's calendar. For each sales call, run the meeting-prep skill, send the briefing doc, and save it to memory.
Test it: docker exec openclaw openclaw agent --message "Prep me for the call with [Name] tomorrow" --deliver
Part 7: Training and Optimization Schedule
Getting everything set up is just the beginning. Here is how to make the system smarter over time.
Week 1: Training Phase
Focus on teaching the AI your voice. Review every draft it writes and give feedback. If an email sounds too formal, tell it to be more casual. If it is missing context about your industry, add more detail to the USER file. Run the inbox monitoring and review the categorizations. This calibration phase is important.
Week 2: Automation Phase
By the second week, the inbox autopilot should be running smoothly. Start using the outreach templates for real prospects. Let the AI draft the first version of every email, then edit it before sending. Enable the follow-up sequences and monitor the first few to make sure the tone feels right.
Week 3: Optimization
Look at which email templates get the best response rates. Check which follow-up sequences are working. Review the meeting prep docs and add any recurring topics or objections. Refine the SOUL file if the AI consistently makes the same kind of mistake.
Ongoing: Continuous Improvement
Spend ten minutes a week reviewing:
- Are the inbox categorizations accurate?
- Are the outreach drafts getting closer to what you would write?
- Are the meeting prep docs covering the right information?
- Are there new patterns or templates you should add?
The compound effect is real — after a few months, the AI is drafting messages that need minimal editing, flagging the right emails as urgent, and building prep docs that cover exactly what you need.
What This Costs vs. The Alternative
- Cloud server: $5 to $10 per month
- Anthropic API usage: $15 to $25 per month (depends on volume)
- Brave Search: Free tier covers most use cases
- Gmail: Whatever you are already paying for Google Workspace
Total: $20 to $35 per month.
Compare that to:
- Outreach.io, SalesLoft, or similar: $100 to $200 per user per month
- AI SDR platforms (11x, Artisan, etc.): $500 to $2,000 per month
- Hiring an actual SDR: $4,000 to $6,000 per month plus benefits
Why Ownership Matters
You own everything in this setup. The AI runs on your server. Your emails stay on your server. Your prospect data stays on your server. Your outreach templates and sales playbooks are files on your machine, not locked inside someone else's platform.
If you decide to stop using OpenClaw tomorrow, you keep all of it. The templates, the email history, the prospect research — it is all yours. Try doing that with a commercial AI sales tool.
Troubleshooting Quick Reference
Token expired / invalid_grant: Re-authenticate with the two-step gog auth flow. This happens every 7 days in testing mode.
Keyring password prompt: Set the GOG_KEYRING_PASSWORD environment variable in your container.
Redirect URI mismatch: Recreate your OAuth credentials as "Desktop app" (not "Web application").
API not enabled: Go to Google Cloud Console and enable the Gmail, Calendar, People, and Drive APIs.
Heartbeat not running: Check with docker exec openclaw openclaw config get agents.defaults.heartbeat. Manually trigger with docker exec openclaw openclaw agent --message "Read HEARTBEAT.md and execute" --deliver
Emails not sending: Test gog directly with gog gmail send --to YOUR_EMAIL --subject "Test" --body "Testing"
Download Everything
Grab all the files you need and drop them into your OpenClaw workspace:
- SOUL Sales Template — AI personality, sales philosophy, output formatting
- Heartbeat Template — inbox monitoring and daily summary config
- Email Templates — cold emails, follow-ups, response handlers, personalization checklist
- Setup Commands Reference — every command from this guide in one copy-paste file
Getting Help
If you get stuck at any point during setup:
- The OpenClaw documentation covers the technical details of the platform itself
- This guide covers the sales-specific configuration and automation setup
- Email support is available at Ali@studytaco.com for questions about any part of the process
Most people get through the entire setup in an afternoon. The parts that take the most time are the Gmail OAuth configuration (about thirty minutes) and writing your SOUL and USER files (about an hour, since you want to be thorough). Everything else is straightforward.
Start Building
The best time to set this up is before you need it. Once the system is running, it works in the background every day — monitoring your inbox, finding opportunities, and keeping conversations alive. The longer it runs, the better it gets.
You do not need permission from IT, approval from your board, or a six-figure budget. You need a $5 server, an afternoon, and the willingness to try something different. The companies that build their own sales infrastructure instead of renting it from vendors will have a structural advantage that compounds over time.
The templates are above. The commands are ready to copy. Go build it.