Build a Cold Outreach Pipeline with OpenClaw (No Code, No Freelancers)

The SOUL.md Configuration

Here's a simplified version of the prompt that drives this pipeline:

You are an outreach assistant for [Agency Name], a [type] agency
that helps [target market] with [services].

Your job is to research prospects, write personalised cold emails,
and manage follow-up sequences.

Research guidelines:
- Focus on companies with 5-50 employees
- Look for pain signals: hiring posts, tool complaints, growth stages
- Find the decision-maker (founder, CEO, VP Marketing, Head of Ops)
- Note 1-2 personalisation angles per prospect

Email guidelines:
- Keep emails under 100 words
- Lead with the personalisation angle, not your pitch
- One clear CTA per email
- No jargon, no buzzwords, no "I hope this finds you well"
- Sound like a human, not a sales tool

Follow-up rules:
- Wait 3 business days between touches
- Maximum 3 follow-ups per prospect
- Each follow-up adds new value (case study, insight, social proof)
- If they say no, respect it immediately

Tone: Direct, helpful, slightly informal. Like a smart colleague
recommending something, not a salesperson pushing a quota.

Why Self-Hosting This Is a Nightmare

Cold outreach automation needs to run reliably, every day, without supervision. Self-hosting OpenClaw for this use case means:

Uptime is critical. If your server goes down for 4 hours, follow-up emails don't send on schedule. Prospects who replied don't get routed. Your pipeline stalls — and you don't know until someone checks.

Credential management is complex. You're storing email API keys, enrichment tool credentials, and CRM tokens on a VPS. That's a security risk you're personally responsible for.

Memory management matters. The agent needs to remember which prospects it's contacted, what stage they're in, and what personalisation angles it used. Self-hosted OpenClaw's memory can be fragile — a misconfigured restart can wipe conversation history.

Token costs add up silently. Researching 50 prospects per day burns through API tokens. Without cost tracking, you might not notice until your Anthropic bill arrives.

On Clawfleet, all of this is handled. The agent runs in an isolated container with health checks every 60 seconds, automatic restarts, encrypted credential storage, persistent memory across restarts, and a cost dashboard that shows exactly what you're spending per day.

The Math

Traditional cold outreach stack:

• Apollo: $99/month
• Instantly: $97/month
• Clay: $149/month
• VA for personalisation: $500–$1,500/month
• Total: $845–$1,845/month

OpenClaw on Clawfleet:

• Clawfleet instance: $9.99–$99/month (depending on plan)
• API costs (Claude Haiku for research, Sonnet for writing): ~$30–$80/month
• Total: $40–$180/month

That's an 80–90% cost reduction. And unlike the VA, the agent doesn't call in sick, doesn't need training on your new service offering, and sends follow-ups at exactly the right time.

Getting Started

Step 1: Sign up for Clawfleet and create your instance. Takes 60 seconds.

Step 2: Configure your SOUL.md with your agency's positioning, target market, and email tone.

Step 3: Install the Web Search, Email Sender, and Scheduler skills.

Step 4: Start small — research and email 10 prospects manually first, then let the agent handle follow-ups. Once you trust the output, expand to full automation.

Step 5: Set up the Propose & Approve workflow. The agent drafts emails and waits for your approval before sending. After 3 consecutive approvals, it earns autonomy for that template pattern.

The agencies getting the best results start with 20 prospects per week and scale to 100+ once the messaging is dialled in. The constraint isn't the tool — it's the quality of your ICP definition and messaging framework.

Stop paying $1,500/month for tools and VAs to do what one AI agent handles. Start your $1 trial on Clawfleet and build your outreach pipeline today.