MCP Cookbook: 12 Marketing Workflows You Can Steal (With Prompts)

Workflow 1: Monday morning marketing pulse (10 min)

Goal: Know what changed over the weekend across all marketing channels before your first meeting.

Prompts:

1. "Pull GA4 sessions for Saturday + Sunday vs prior weekend.
Show top 10 pages by session change."

2. "Pull Google Ads + Meta Ads spend + conversions for the weekend
vs prior weekend. Flag anything moving more than 30%."

3. "Pull GSC clicks Sat + Sun vs prior weekend. Top 5 winners,
top 5 losers."

4. "Based on above, give me top 3 things I should know before
my 10am meeting."

Saves: 45-60 min of manual platform-hopping. Cost: ~$0.20 in API.

Workflow 2: Friday weekly audit (15 min)

Goal: Identify everything that needs action before week's end.

1. "Run the wasted Google Ads spend prompt (keywords with $50+ spend,
zero conversions, last 7 days)."

2. "Pull Meta ad sets stuck in Learning Limited >14 days."

3. "List GSC pages where impressions are up 20%+ but CTR is below 1%."

4. "Cross-reference: which client accounts (if multi-client) had
ROAS drop >25% week-over-week?"

5. "Generate a 5-bullet priority list for next week based on
the above."

Saves: 2-3 hours of manual audit. Cost: ~$0.50.

Workflow 3: Client renewal prep (20 min)

Goal: Before a quarterly renewal conversation, have the data ready that justifies your fee.

1. "Pull last 90 days vs prior 90 days for client X: total sessions,
conversions, conversion value, spend, ROAS."

2. "Identify the 3 most impactful changes we made for this client.
Tie each to measurable results."

3. "List the 3 issues we caught early that would have cost the client
money if unaddressed."

4. "Draft a 1-page renewal summary in the agency's voice."

Saves: 3-4 hours of pre-meeting data assembly.

Workflow 4: Weekly client report (4 min/client)

1. "Switch context to client X."

2. "Pull last 7 days vs prior 7 days: GA4 sessions, Google Ads spend
+ conversions, Meta Ads spend + conversions, GSC clicks."

3. "Draft a weekly summary in the agency voice (load voice profile).
Format: Top win, top concern, recommended action, one chart description."

4. "Convert to email-ready format. Subject line + body."

Saves: 25-30 min per client. For 20 clients: ~10 hours/week.

Workflow 5: Attribution debug session (30 min)

Goal: When numbers don't add up between platforms.

1. "Pull last 30 days conversions from: GA4, Google Ads, Meta Ads.
Show the discrepancies."

2. "For Google Ads conversions, show breakdown by conversion action
and attribution model."

3. "For Meta conversions, show breakdown by attribution window
(7-day click vs 1-day click vs view-through)."

4. "Identify the largest source of overlap and recommend a single-
source-of-truth approach."

Saves: 2-4 hours of manual reconciliation.

Workflow 6: Meta creative fatigue scan (8 min)

1. "List Meta ads where frequency > 5 AND CTR has dropped > 25%
in the last 7 days vs prior 7 days."

2. "For each, pull the creative thumbnail and identify what made
it work originally."

3. "Recommend: pause, refresh, or test variation."

4. "Suggest 2-3 variations for the top 3 fatigued creatives."

Saves: ~60 min of manual ad-by-ad review.

Workflow 7: New campaign launch check (15 min)

1. "Pull current performance of all campaigns running for keyword
'[target keyword]' in our category."

2. "Show competitor ad copy variations from Google Ads Transparency
Center for the same keyword."

3. "Draft 3 new ad copy variations for the target keyword based on
what's working in the category."

4. "Recommend bid range based on current competitive CPC for this
keyword."

Saves: 1-2 hours of research before launching.

Workflow 8: Month-end budget reallocation (10 min)

1. "Pull last 30 days spend + conversions + ROAS for each Google Ads
+ Meta Ads campaign."

2. "Sort by marginal ROAS (last 7 days, not full 30) ascending."

3. "If I had $5K of new budget for next month, which campaigns
deserve it based on marginal ROAS?"

4. "What's the expected incremental conversion impact?"

Workflow 9: Content gap analysis (12 min)

1. "From GSC, pull queries where I have impressions but rank below
position 10. Filter to commercial-intent terms (e.g., contain words
like 'pricing', 'vs', 'best', 'alternative')."

2. "Show top 20 sorted by impression count."

3. "For each, suggest the page type that would rank: pricing page,
comparison page, vs-page, listicle, or blog post."

4. "Prioritize the list by 'effort to rank' (low DA needed first)."

Workflow 10: Competitor analysis sprint (25 min)

1. "Ask Claude (without MCP): list top 5 competitors in [my category]."

2. "For each, pull their estimated organic traffic via [external tool if
connected] OR ask 'what are their top 5 ranking pages'?"

3. "From GSC, identify queries where their domain appears alongside
mine in SERPs. Top 15."

4. "Recommend 3 specific content angles to win on the overlap queries."

Workflow 11: Pricing-page conversion debug (15 min)

1. "Pull GA4 sessions on /pricing page last 30 days. Show: total sessions,
average time on page, bounce rate, sessions that proceeded to /auth.html."

2. "Break down by traffic source — which source converts pricing-page
visitors best?"

3. "Compare current month vs prior 3 months — has conversion rate
on pricing changed?"

4. "Suggest 3 A/B test ideas based on data patterns observed."

Workflow 12: AI citation tracking (8 min weekly)

1. "Pull GA4 sessions from chatgpt.com, claude.ai, perplexity.ai,
gemini.google.com last 7 days vs prior 7 days. Show % change per source."

2. "For each AI source, show top 3 landing pages."

3. "Identify which content pieces are getting AI citation lift —
these are templates to replicate."

4. "Flag any AI source where traffic declined >30% — may indicate
citation pool refresh."

This single workflow has driven most of our content strategy decisions in 2026. If a piece of content lands AI citations, you replicate the format. If a piece stops getting cited, you investigate why before assuming it's lost forever.

How to build your own workflows

• Start with a pain point — a recurring task that eats time. Workflow-design beats prompt-design.
• Sequence 3-5 prompts — most useful workflows are multi-step. Single prompts give shallow answers.
• Save the prompt sequence — Claude lets you save custom commands; build a personal library.
• Iterate based on output quality — if a workflow's output is weak, the fix is usually a more specific system prompt or better data filter, not a different model.
• Share workflows with your team — the highest-ROI MCP work in agencies happens when 3+ people share and refine the same workflow library.

Frequently Asked Questions