Draft + humanize
Claude writes it. Walter rewrites it so it reads like a human wrote it. One step, not two tools.
Step 01 → 02
Walter is an MCP server you add to Claude. Once it’s connected, Claude can humanize a draft, run AI detection, and lock your exact target phrases — all in one chat. This site is the unofficial recipe book: 65 copy-paste prompts across 12 categories and 12 Claude project Skills collected from teams running it at agency scale.
Free MCP install. Free prompt library. No signup required to read anything here.
Google's March 2026 update hit sites publishing AI content without editorial oversight. Detection tools are better than ever. Clients are asking harder questions about where the content comes from.
The agencies still standing are the ones that figured out how to add a quality layer to AI output without slowing down production. But most teams are doing it the hard way — rewriting by hand, running content through three different tools, or hiring editors to clean up what the AI produced.
That workflow breaks at scale. And when it breaks, you lose rankings, lose clients, or both.
Walter is a humanization engine that runs inside Claude as a connector. You write one prompt; Claude drafts, Walter humanizes, detection runs, keywords are confirmed. The prompt pack gives you 50 of these workflows, organized by use case.
Claude writes it. Walter rewrites it so it reads like a human wrote it. One step, not two tools.
AI detection runs automatically. You see the score before you publish. No separate tab, no separate tool.
You specify which phrases to protect. Walter locks them before humanizing so they survive intact. 100% preservation rate.
Every prompt runs the full pipeline: draft, humanize, detect, preserve keywords. You don't have to remember what to ask for. The prompts handle it.
Long-form drafts that preserve exact phrases through every humanization pass.
10 prompts →Batch CSV-to-copy flows with brand and SKU lock-ins.
8 prompts →City + service templates that hold the H1 phrase across every page.
7 prompts →Turn one post into a thread, newsletter, and carousel without re-detecting.
5 prompts →Pre-flight checks. Side-by-side detection scores. Phrase coverage reports.
5 prompts →Style-guide-aware briefs and client-ready deliverables in one pass.
5 prompts →Hero, sub, CTA, and FAQ blocks tuned to convert without sounding generated.
5 prompts →Schema explainers, docs, and how-to guides that pass detection at depth.
5 prompts →Threads, captions, and platform-native posts repurposed from one source.
5 prompts →Newsletters, nurture sequences, and product emails with locked tone.
4 prompts →Cold email, link pitch, and broken-link outreach with specific hooks.
3 prompts →Sales one-pagers, LinkedIn DM prospecting, and follow-up scaffolds.
3 prompts →Copy-paste directly into Claude with Walter connected. Each one runs the full pipeline.
Write a 1,200-word blog post targeting [keyword]. Preserve these exact phrases: [phrase 1], [phrase 2], [phrase 3]. Humanize it with Walter in balanced mode, then run AI detection and show me the keyword preservation report.
What happens: Claude drafts the article, Walter humanizes it, detection runs, keywords are verified. You get the finished article plus a score and a keyword confirmation. Under 2 minutes.
Write a 300-word landing page for "[service] in [city]". Preserve the exact phrase "[service] in [city]" in the H1, first paragraph, and last paragraph. Include one mention of [neighborhood]. Humanize with Walter.
What happens: Run it 50 times with different cities and you have 50 unique, humanized landing pages with location keywords intact.
Here is a draft: [paste text]. Humanize it with Walter in balanced mode. Then run AI detection on both the original and humanized versions. Show me a side-by-side comparison with scores for each.
What happens: Your writer submits a draft. You paste it in. You get back a humanized version with a before/after detection comparison. QA in 30 seconds.
Open Claude → Settings → Connectors. Add Walter: mcp-server.walterwrites.ai/mcp
Copy any workflow from the library. Fill in the bracketed placeholders.
Claude drafts, Walter humanizes, detection runs, keywords stay locked. Done.
Works with any Claude plan, including free. Walter's free tier has enough words to test 3–5 full articles.
Install the MCP. Drop in a Skill. Watch every chat draft, humanize, and detect on its own.
Walter installs as an MCP connector in about 30 seconds. Free tier covers your first articles — no card.