The Webprofits Stack (MCPs, APIs, Tools)

Overview

The full inventory of what we run at Webprofits day to day - every MCP, every API, every tool, and the skills layer on top. This is a reference. Bookmark it, come back to it. You don't need to install everything on day one. Start with Tier 1, then layer in Tier 2 and Tier 3 as the work calls for them.

There are three tiers of MCP setup

1. claude.ai connectors (one-click) - OAuth handled by claude.ai. You click Connect, log in, done.
2. 3rd party MCPs (sign up + connect) - separate service account, then connect through Claude Code.
3. Custom MCPs (Cloud Console + OAuth credentials) - your own dev project, API enabled, credentials wired in.

Plus a layer of direct APIs (called via .env keys, no MCP) and the tools and infrastructure (terminals, storage, publishing) that holds it together.

Tier 1 - claude.ai connectors

One-click OAuth through claude.ai. No Cloud Console, no API keys to manage. These are the fastest wins on day one.

Google Drive. File CRUD - create, read, update, delete. Drive MCP creates new Docs and Sheets from scratch fine, but it does NOT edit cells in existing Sheets or do granular Doc edits - for that you need the Tier 3 dedicated MCPs below. Gmail. Search threads, read messages, draft replies, label, organise. We use it inside Claude Code for searching past client emails, drafting responses in our voice from thread history, and applying labels to organise threads.

Google Calendar. Events, scheduling, suggest time, list calendars. We use it to schedule client calls, internal team blocks, and to pull "what's on next week" context into wrap-up sessions.

Slack. Read channels, post messages, manage canvases, search across public + private. We use it for HR updates, hire announcements, ops broadcasts, and pulling thread context into Claude Code (e.g. "summarise the full thread in #client-channel").

Notion. Notes, databases, briefs. Our static-ads-brief database is here. We use it for client briefs, content pipelines, and team documentation that doesn't live in code.

Supabase. Every project's brain. Per-person memory tables, per-client brain tables, plus the central Hub. The startup and wrap-up skills read and write through this MCP. This is the single biggest unlock once configur ed.

Klaviyo. Flows, campaigns, segments, metrics, templates, profile management. The MCP is light on the audit side, so we fall back to the direct API for full deep dives. Day to day, the MCP covers campaign sends, list management, and reporting.

Meta Ads MCP. Read auth and account context. Heavy lifting on Meta writes goes through Pipeboard.

Figma. Design context, screenshots, code-connect mapping, design tokens. We use it for design-to-code work and pulling design specs into Claude Code.

Fathom. Internal-only - meeting transcripts and summaries. We use it for internal call reviews, not for client-facing content.

Context7. Library/framework documentation lookup. Use it whenever a skill is touching a specific library or API - it returns current docs rather than relying on training data. Supermetrics. Cross-platform data aggregation across 150+ marketing sources. We use it when a client needs blended reporting across channels that don't have direct MCPs.

Zapier. Slack + Airtable bridge for workflows that need both.

n8n. Workflow build and deploy. We use it for scheduled automations that need to run outside Claude Code (e.g. nightly Klaviyo syncs, daily CV processing).

Tier 2 - 3rd party MCPs

A separate service account, then connect through Claude Code's connector flow . No Cloud Console needed - the vendor handles auth on their end.

Pipeboard Google Ads / Meta Ads / TikTok. Our primary write path for paid media. Pipeboard.co account, connect via Claude Code connectors, OAuth into each ad platform through Pipeboard. Launches campaigns, adsets, ads, RSAs, sitelinks, keywords, audiences, geo, language targeting, bid adjustments, budgets, pauses, ad copy edits. We don't open Ads Manager unless we have to. Approval discipline matters - "do not change anything" prompt before any write to a live client account, then explicit per-step confirmation.

DataForSEO MCP. Keyword research, SERP scrape, backlink data, content phrase trends, search intent, domain analytics, YouTube SERP, business listings. We use it for client enrichment, SEO audits, content strategy, and keyword research before launching any new search campaign. DataForSEO is a paid account (cheap per query) - basic-auth credentials go in the MCP config.

Tier 3 - custom MCPs (Cloud Console + OAuth)

These exist where the claude.ai connector doesn't exist OR is too limited for the work. Setup is heavier than Tier 1 and 2 but still quick - your own Google Cloud Console project (free tier is fine), the relevant API enabled, OAuth client created, credentials downloaded, MCP config wired in. Tier 3 is optional - only set up what your work calls for. GA4 MCP. No claude.ai equivalent. Requires Google Cloud Console + Analytics Data API + OAuth. We use it for client reporting - sessions, conversions, traffic source breakdowns, ecommerce events, custom dimensions. The reporting skill calls this directly to build monthly dashboards.

GSC MCP (Google Search Console). No claude.ai equivalent. Requires Cloud Console + Search Console API + OAuth. We use it for SEO audits, search performance reports, URL indexing checks, sitemap management, search-by-page-query analysis. Required for any client where SEO is in scope.

GMC MCP (Google Merchant Center). No claude.ai equivalent - a custom MCP. Requires Cloud Console + Content API v2.1 + OAuth. We use it for product feed health, account-level diagnostics across all our ecommerce clients, and Shopping ad foundations.

Google Sheets MCP. claude.ai Drive can read sheets as files, but it cannot edit cells. This MCP handles cell-level operations - update_cells, batch updates, sheet metadata. Requires Cloud Console + Sheets API + OAuth. We use it when a skill needs to write into an existing dashboard sheet (e.g. KPI auto-updates).

Google Docs MCP. Same pattern. claude.ai Drive can create a doc and read content, but cannot insert text at a specific position or apply formatting in an existing doc. This MCP handles granular edits. Requires Cloud Console + Docs API + OAuth. We use it for proposal updates, retainer doc edits, and content workflows that need formatted output.

LinkedIn Ads MCP. LinkedIn doesn't go through Google Cloud Console - it has its own developer portal, but the setup pattern is identical. LinkedIn Developer Portal app + Marketing API approval (LinkedIn approves the developer access separately, can take days) + OAuth. We use it for LinkedIn campaign creation, creative builds, lead form management, conversion tracking. Lead gen accounts only.

Direct APIs (via .env keys)

Not MCPs - these are direct API calls from skills using credentials stored in per-client

.env files (gitignored).

Anthropic API. The core LLM behind every prompt, every skill, every agent. Webprofits has the top-tier subscription for the whole team plus burst capacity.

OpenAI API. Embeddings only (text-embedding-3-small). Used for semantic search across past sessions in Supabase memory.

Gemini API (2.5 Flash Image + Pro Vision). All image generation and image analysis. Banner pipeline, creative mockups, ad asset analysis, vision-on-screenshot tasks.

Apify API. Web scraping fallback when Claude's WebFetch can't render a JS-heavy site. Also our Reddit scraper for client enrichment, and our Facebook comments scraper for ad engagement analysis.

Klaviyo API. Per-client API key for full audit access (the MCP is light). Stored in clients/{brand}/.env. Used by the klaviyo-audit skill.

Shopify Admin API. Per-client custom app, full storefront and order access. Used for ecommerce reporting, product feed work, order-data analysis.

Stripe Reporting API. Monthly subscription and financial reports.

Air.inc API. Digital asset management. Banner library upload + naming-convention tagging so post-launch analysis can query by hook / persona / angle.

Triple Whale API. Ecommerce attribution overlay on top of Shopify data. Used for blended attribution reporting on ecomm clients.

Dropbox API. Asset upload pipeline for clients where assets need to land in a shared Dropbox folder (e.g. designer handoff).

Tools + infrastructure

Claude Code CLI. The primary work environment - terminal only, not browser. The course exists because the browser is 5% of what's possible. Warp (Mac + Windows). Terminal app with tab-based multi-session. The reason this matters - parallel Claude Code sessions across tabs is what unlocks speed.

Superset (Mac only). Alternative terminal with multi-session support. Windows users go Warp.

Cursor / VS Code. Developer fallback when the work is code-heavy. Claude Code runs inside both.

Supabase. Database layer - per-person memory, per-client brain, central Hub.

GitHub. Shared skills repo (webprofits org) + per-client repos + GitHub Pages publishing layer. All live dashboards deploy here.

OneDrive / Google Drive / Dropbox. File sync across machines. Pick one (we use all three for redundancy). Working folders live in Dropbox by default; GitHub captures version-controlled work; OneDrive for the rest.

1Password. Rotating per-client API key vault. Source of truth for every key before they get copied into .env files.

.env files (gitignored). Per-client secrets at runtime. Never committed.

GitHub Pages (Actions workflow). Publishing layer for all live client dashboards, reports, audits.

StatiCrypt. Password-protects client dashboards. HMAC mode for stronger gates, shared-password mode for simpler gates.

How to use this list

Tier 1 and Tier 2 are for everyone. Get all the claude.ai connectors live, then Pipeboard and DataForSEO. That's the baseline stack - it covers the daily work for marketers, founders, growth leads, and agency owners. Tier 3 is optional. Set them up only when the work calls for it. Don't set up GA4 until you need GA4 reporting. Don't set up GMC until you have an ecomm client with a feed. Don't set up LinkedIn until lead gen is in scope.

Direct APIs go into .env files per client as you go - sign in, generate key, paste, restart session.

Exercises

1. Review the concepts covered in this lesson: The Webprofits Stack (MCPs, APIs, Tools).
2. Write down your key takeaway from this lesson.
3. Practice running any commands or prompts mentioned above inside your terminal.