How to draft link-building outreach with Claude that actually gets replies.

Claude Projects is the most powerful no-code method for learning how to draft link-building outreach with Claude at scale. You store your ideal customer profile, brand voice guide, outreach rules (under 150 words, no salesy language, reference one recent article), and link-worthy asset description as persistent project context. Every conversation then inherits those rules automatically, so you never re-paste your brief.

The workflow is straightforward: paste a batch of enriched prospect rows — name, site, niche, recent post title, content gap — and prompt Claude to generate a unique opening line and full email for each. Claude’s 200K-token context window handles 50+ prospects in a single pass. Review, approve, and copy the output into Instantly or Smartlead. Teams using this approach cut drafting time from 30 minutes to under 3 minutes per prospect and report reply rates of 15–22% on well-researched lists. Pair it with a tool like Ahrefs or Semrush to identify the content gap your asset fills before you prompt. For a deeper look at connecting Claude to your broader marketing stack, see our guide on how to connect Claude to Google and Meta Ads via MCP.

Clay is an enrichment platform that scrapes LinkedIn posts, company news, recent funding rounds, and website changes for each prospect. When you pipe those enriched columns into a Claude prompt template inside Clay, the result is outreach that references the prospect’s actual LinkedIn activity or a specific piece of content they published last week — the kind of signal that makes an editor think you actually read their site.

The setup involves building a Clay table, adding enrichment waterfalls for each data point you want Claude to reference, and writing a prompt column that pulls those variables in. The output drops into your sequencer automatically. The main constraint is cost: heavy enrichment burns Clay credits fast, so this method suits teams running 200+ targeted prospects per week where the reply-rate lift justifies the spend. If you are earlier in your SEO journey, our post on using Claude for SEO covers how to build the underlying asset strategy that makes your outreach worth personalizing in the first place.

Claude Code is Anthropic’s command-line developer tool. A basic script reads a CSV of enriched prospects, sends each row to the Claude API using a carefully engineered template prompt, and writes personalized email drafts to an output file. A 500-prospect batch typically runs in under 10 minutes and costs less than $20 in API credits — a fraction of what manual writing or a human copywriter would cost.

The production-ready prompt structure that works best follows a clear pattern: define the role (“You are an expert B2B outreach copywriter”), insert the full brand voice guide, pass in the prospect variables (company, role, recent post, content gap, trigger event), specify constraints (under 150 words, reference one specific article, low-friction CTA like “Would love to contribute a piece on X that fits your readership”), and set tone. Review the output file before pushing to your sequencer — supervised personalization at speed, not blind automation. For the technical setup, Anthropic’s documentation covers the API authentication in detail.

Instantly and Smartlead are the two dominant cold-email sequencers used by link-building agencies. Neither has a native Claude integration, but the handoff workflow is clean: use Claude Projects or the API to generate send-ready email copy for each sequence step (initial outreach, follow-up 1, follow-up 2, break-up), then paste the drafts into the sequencer’s step editor with your standard variable fields mapped in.

The key insight for link-building specifically is to generate variation copy for each follow-up inside the same Claude session, so tone stays consistent across the sequence. A three-step sequence that opens with a content-gap angle, follows up with a social-proof data point, and closes with a low-friction “no worries if not” message typically outperforms a single-email blast by 40–60% in reply rate. Claude handles all three steps in one prompt pass. Export as plain text and import into sequence steps — no API required.

Model Context Protocol (MCP) lets Claude connect directly to external tools and data sources. A browser-automation MCP server — built on Playwright or a service like Browserbase — lets Claude visit each target URL, read the actual page content, identify the most recent article, extract the author’s name, and then draft an outreach email that references what was actually published that morning. This is the gold standard for personalization depth.

The practical workflow: pass Claude a list of target URLs. The MCP server fetches each page. Claude reads the content, identifies a relevant content gap your asset fills, and writes a 120–150 word outreach email per site. Output is written to a structured JSON or CSV file ready for your sequencer. Throttle requests to avoid rate limits — one request per 3–5 seconds is safe for most sites. This approach is explored in depth in our guide on connecting Claude to external tools via MCP.

Apollo solves the prospecting half of link-building: finding verified email addresses for editors, content managers, and site owners. Once you have your contact list, Claude writes the copy — paste the Apollo-enriched prospect data (company, role, recent activity if available) into a Claude Project session and generate personalized drafts for each sequence step. Then paste the approved copy into Apollo’s sequence editor, mapping Claude’s output to the step fields.

The limitation is that Apollo’s enrichment skews toward sales data (company revenue, tech stack, funding stage) rather than editorial signals like recent post topics or content gaps. For pure link-building personalization, Clay enrichment feeds Claude better raw material. But if your team already uses Apollo for outbound sales, adding Claude-drafted copy to your link-building sequences costs almost nothing extra and meaningfully improves reply rates over generic templates.

A structured template library stored inside a Claude Project is the most accessible entry point. Keep five to seven proven outreach frameworks as project documents: cold outreach leading with a content gap, trigger-event outreach (referencing a recent site update or award), broken-link outreach, resource-page outreach, and a follow-up sequence. Each template has placeholder variables for the prospect name, site niche, specific article reference, and your link-worthy asset URL.

To draft a new email, open the project, paste the prospect’s details, and tell Claude which template to use. Claude fills the variables with genuinely personalized language rather than literal placeholder text, keeping output under 150 words and ending with a soft CTA. The process takes under three minutes per email. At 10–20 emails per day this is entirely manageable; beyond that, move to the API batch method. For more on building Claude workflows for SEO, see our post on Claude for SEO.

This workflow pairs Claude with a structured Notion database that tracks every prospect from discovery through placement. Each Notion row holds the target site URL, editor name, niche, content gap identified, draft status, send date, reply status, and link placement confirmation. To generate outreach, you grab the relevant columns from the row, paste them into a Claude chat (or API call triggered via Notion automation), and write the approved draft back into the row.

The main advantage is auditability. When a link goes live six weeks later, you can trace exactly which outreach version, follow-up sequence, and Claude prompt generated the placement. That institutional knowledge compounds — over time you can identify which template structures, content-gap angles, and CTAs drive placements for specific site categories. Teams that have run this workflow for three months or more typically see their best-performing prompt patterns drive 30–40% of all placements, which they then prioritize at scale.

For a handful of priority targets — a DA 70+ industry publication, a newsletter with 50K subscribers, or an editorial placement that would move your topical authority significantly — a fully manual Claude session produces the highest-quality drafts. Research the target thoroughly before opening Claude: read their last five published pieces, identify a specific angle in your asset that fills a gap in their coverage, and note the editor’s writing style and preferred tone.

Then open a Claude session (use Projects if available for persistent brand voice), provide the full research brief, and ask Claude to draft an email that opens with a specific reference to their most recent relevant piece, connects that to the gap your asset fills, and closes with a single low-friction ask. Review, edit for personal voice, and send. This method has the highest ceiling for reply quality but the lowest ceiling for volume — treat it as your sniper approach alongside a higher-volume automated method for the rest of your list.