The problem
Outreach is either personal or scalable — rarely both
Manual outreach that actually earns links is painstaking: research each prospect, verify the domain is worth it, find the right contact, write something they'll actually reply to, then chase. Scale it with templates and reply rates collapse — plus you risk your sender reputation.
This agent does the research and personalisation at volume, and keeps a human in control of every send.
Primary objective
Earn quality links at scale without spraying templates or losing the personal touch.
How it works
Prospect → qualify → personalise → approve → follow up
Prospect
Find topically relevant sites
→
Qualify
DA, traffic, spam & fit checks
→
Personalise
Email grounded in their content
→
Approve
Human signs off each send
→
Follow up
Sequenced, reply-aware nudges
What it does
- ✓ Topical prospect discovery
- ✓ Domain quality & spam scoring
- ✓ Contact discovery & verification
- ✓ Content-grounded personalisation
- ✓ Reply-aware follow-up sequences
- ✓ Won-link tracking & reporting
Built with
- ✓ Tool-using agent (Claude + MCP)
- ✓ Backlink & domain-metrics APIs
- ✓ Email verification & deliverability
- ✓ CRM / inbox integration
- ✓ Approval queue with edit-in-place
Guardrails
- ✓ No send without human approval
- ✓ Volume caps to protect reputation
- ✓ Suppression & do-not-contact lists
- ✓ CAN-SPAM / GDPR-aware templates
- ✓ Full message audit log
Under the hood
Personalisation grounded in the prospect's own content
The agent reads a recent, relevant page from the prospect and references it specifically — every draft lands in an approval queue, never the outbox.
outreach/personalise.py
PYTHON
from anthropic import Anthropic from approvals import queue_for_review client = Anthropic() def draft_outreach(prospect: dict, our_asset: str) -> dict: if prospect["score"] < 0.6 or prospect["suppressed"]: return {"skipped": True} # qualify before we ever draft prompt = f"""Write a short, genuine outreach email. Their recent article: {prospect['recent_page_summary']} Reference it specifically (no flattery filler). Our relevant resource: {our_asset} One clear, low-friction ask. Under 120 words. No hype.""" msg = client.messages.create( model="claude-sonnet-4-20250514", max_tokens=400, temperature=0.5, messages=[{"role": "user", "content": prompt}], ) # Never auto-send — a human approves or edits first return queue_for_review(prospect, msg.content[0].text)
Outcomes
5×
More qualified prospects/week
Higher
Reply rate vs templates
Human
Approval on every send
Safe
Sender reputation protected
Scale outreach without going spammy
Run it standalone, or feed it link-worthy assets from the Content Engine.
More SEO & Marketing solutions