Training

AI Tool Setup

Needs review — This SOP contains our content but has not been verified by Nick. Treat as a working draft until marked Live.

SOP: AI Tool Setup Guide

Last Updated: 2026-03-26 Version: 1.0

Purpose

One-time setup guide for Claude Code, Cloudflare Workers, API tokens, and per-client Claude projects. Complete this during your first week. These tools are your daily drivers for content creation, data analysis, and technical implementation.

Estimated Time: 2-3 hours Prerequisites: Completion of 11-seo-foundation-training.md

Section 1: Claude Code Setup

Claude Code is your AI command center. Generates content, analyzes data, creates schemas. Used daily.

Step 1.1: Install Claude Code via npm

Open terminal and install:

npm install -g @anthropic-ai/claude-code

Troubleshooting:

“command not found: npm” → Install Node.js first: https://nodejs.org/ (LTS recommended)
Permission errors → Try: sudo npm install -g @anthropic-ai/claude-code
Verify: claude-code --version

Step 1.2: Get Your Anthropic API Key

Go to https://console.anthropic.com/
Sign up or log in with work email
Go to Settings > API Keys
Click Create New API Key
Copy the key (you won’t see it again)
Store securely (next step)

Security Note: This key is sensitive. Never share it, never commit to git, never post in Slack.

Step 1.3: Set Your API Key in Shell Profile

For Mac/Linux (zsh or bash):

nano ~/.zshrc   # or ~/.bashrc

# Add this line at the end:
export ANTHROPIC_API_KEY="sk-ant-xxxxxxxxxxxxx"

# Save (Ctrl+O, Enter, Ctrl+X)

# Reload:
source ~/.zshrc   # or ~/.bashrc

For Windows (PowerShell):

[System.Environment]::SetEnvironmentVariable('ANTHROPIC_API_KEY', 'sk-ant-xxxxxxxxxxxxx', 'User')

Verify:

echo $ANTHROPIC_API_KEY

Should return your key (truncated). If blank, setup failed.

Step 1.4: Create Local Directory Structure

mkdir -p ~/tekton-growth/clients
mkdir -p ~/tekton-growth/templates
mkdir -p ~/tekton-growth/scripts
mkdir -p ~/tekton-growth/output

Step 1.5: Test Your Setup

claude-code "Write a 100-word blog post about local SEO for a plumbing business in Austin, TX. Include keywords 'emergency plumber austin' and 'plumbing repair near me'."

Expected output: A blog post draft in your terminal

If it fails:

Check API key: echo $ANTHROPIC_API_KEY
Check internet connection
Verify API credits (console.anthropic.com)
Slack #tech-support if stuck

Section 2: Per-Client Claude Projects

Each client gets their own Claude project on Claude.ai. Keeps context separated, enables faster, accurate prompting.

Step 2.1: Create a New Claude Project

Go to https://claude.ai/
Click Projects (left sidebar)
Click Create New Project
Name it: [Client Name] - [Year] (e.g., “Austin Plumbing Co - 2026”)
Click Create

Step 2.2: Upload Project Instructions

Customize and upload this template to each project:

Template: Client Project Instructions

You are an SEO specialist working on [CLIENT NAME]'s local SEO campaign.

---
BUSINESS PROFILE
Business Name: [Full legal name]
Website: [URL]
Service Areas: [Cities/regions served]
Primary Service: [Main offering]
Years in Business: [X]
Unique Value Prop: [What sets them apart]

---
AUDIENCE AVATAR
Primary Customer: [Name, age, pain point, intent]
Search Intent: [What are they looking for]
Decision Drivers: [Cost, speed, reputation, expertise, location]

---
TARGET KEYWORDS (Priority Order)
1. [Keyword 1] – Volume [X], Difficulty [X/10]
2. [Keyword 2] – Volume [X], Difficulty [X/10]
[Up to 15 total]

---
BRAND VOICE
Tone: [Professional, casual, expert, friendly]
Language Style: [Simple explanations, technical, storytelling]
Example content: [URL of best existing page]

---
WEBSITE PLATFORM
Platform: [WordPress, Squarespace, GHL, other]
Host/Performance: [Strengths and constraints]
Editing Access: [Who can edit, any limitations]

---
CONTENT GUIDELINES
Content types: [Blogs, service pages, FAQs, GBP posts, etc.]
Publishing cadence: [X posts per month]
Content length: [500w, 1500w, 3000w]

---
TECHNICAL DETAILS
GBP Status: [Verified? Categories optimized? Review count]
Mobile: [Good, needs work, terrible]
Site speed: [Good, acceptable, needs work]
Current schema: [What's already implemented]
Analytics: [GA4 set up? GHL connected]

---
COMPETITOR INTEL
Top 3 competitors: [Names and domains]
Competitor advantage: [What they're doing better]
Opportunity gaps: [What they're missing]

---
PREVIOUS WORK
Best performing content: [URL and why]
Worst performing content: [URL and what failed]
Already tried: [Campaigns, keywords, tactics]

---
END INSTRUCTIONS

You have all context needed to create SEO content and strategy for this client.
When I ask to "generate a blog post" or "analyze data," apply all context above.
Always confirm details match before finalizing output.

Step 2.3: Upload Supporting Documents

For each project, upload (if available):

Brand Guidelines PDF (logo, colors, fonts, tone)
Keyword Research spreadsheet (list with volume, difficulty, intent)
Competitor Analysis (top 3 competitors’ backlinks, content profile)
Audience Avatar Document (detailed customer persona)
Content Audit (best and worst performing pages)
GBP Audit (current state, categories, photos count)

How to upload:

Open the project
Click Upload (in chat interface)
Select file (PDF, .xlsx, .txt, .md supported)
Reference in prompts

Step 2.4: Test the Project

Send a test prompt:

Based on my business profile and audience, write 3 blog post topic ideas that would resonate with my primary customer avatar. Rank by search volume and difficulty. Explain why each would work.

Expected output: 3 custom topic ideas tailored to their business

If it’s generic: Update project instructions and try again

Section 3: Cloudflare Workers Setup

Cloudflare Workers is where we inject schema, handle redirects, and optimize performance without touching client site code.

Step 3.1: Access Tekton Growth Cloudflare Account

Manager sends you account access link (email or 1Password)
Accept invitation, set your password
Log in to https://dash.cloudflare.com/
You should see “Tekton Growth” account in top-left dropdown
Ask manager for “Editor” access for your first month

Step 3.2: Install Wrangler CLI

npm install -g wrangler

Verify:

wrangler --version

Step 3.3: Authenticate Wrangler

wrangler login

This opens a browser. Click “Authorize Wrangler.” You’ll see confirmation in terminal.

Step 3.4: Standard Project Structure

mkdir ~/tekton-growth/workers/schema-injector
cd ~/tekton-growth/workers/schema-injector

Step 3.5: Create a Schema Injector Worker

Use Claude Code to generate production-ready Workers:

claude-code "Create a Cloudflare Worker that injects LocalBusiness schema for [CLIENT NAME].
The business is: [NAME], located at [ADDRESS], phone [PHONE], website [URL].
Include FAQSchema for these FAQs: [FAQ list].
Output the complete src/index.js file ready to deploy."

Copy output into src/index.js.

Step 3.6: Configure wrangler.toml

Edit wrangler.toml:

name = "schema-injector-client-name"
main = "src/index.js"
compatibility_date = "2026-03-01"

[env.staging]
routes = [
  { pattern = "staging.clientdomain.com/*", zone_name = "clientdomain.com" }
]

[env.production]
routes = [
  { pattern = "clientdomain.com/*", zone_name = "clientdomain.com" },
  { pattern = "www.clientdomain.com/*", zone_name = "clientdomain.com" }
]

Step 3.7: Test Locally

wrangler dev

Usually http://localhost:8787. Visit to see Worker in action.

Expected output: Client’s website with schema injected

Debugging:

Check browser console (F12) for errors
Check terminal for Worker errors
Verify schema in HTML source (Ctrl+U, search ld+json)

Step 3.8: Deploy to Staging

wrangler deploy --env staging

Test on the staging URL.

Step 3.9: Deploy to Production

After staging verification:

wrangler deploy --env production

Step 3.10: Post-Deployment Verification

Visit client’s live site
View page source (Ctrl+U)
Search for ld+json
Confirm schema appears with correct data
Test in Google Rich Results: https://search.google.com/test/rich-results

Section 4: Security & Best Practices

API Key Security

DO:

Store in environment variables (never in code)
Use 1Password for backup storage
Rotate keys quarterly
Use separate keys for dev/prod if possible

DON’T:

Commit API keys to git (even private repos)
Share keys in Slack, email, or chat
Use one key for multiple team members
Leave keys in terminal history

If Exposed:

Immediately regenerate key in Anthropic Console
Tell your manager
Check API usage for suspicious activity
Update all saved keys

Client Data Security