I Replaced My SEO Agency With Claude and Four APIs

I Replaced My SEO Agency With Claude and Four APIs

The invoice landed in my inbox on a Thursday morning. $4,200 for the month. My SEO agency had published eight blog posts, scheduled some social media content, and sent me a report with graphs that looked impressive but said nothing I couldn't have figured out by glancing at Google Search Console for ten minutes.

I'd been paying this agency for seven months. The traffic needle had barely moved. Not because they were bad at their jobs — they weren't. The content was fine. The keyword research was competent. But "competent" doesn't cut it when you're competing against thousands of sites publishing daily, and "fine" content doesn't rank when Google's algorithm rewards genuine expertise over well-structured filler.

That Thursday, I cancelled the contract. Not because I'd found a cheaper agency. Because I'd spent the previous weekend building something that made agencies optional for my workflow entirely.

Four tools. One orchestrator. Zero monthly retainers.

Claude Code handles the strategy and analysis — the brain. Arvow writes and publishes SEO-optimized articles directly to my sites. Blotato takes each published post and distributes it across nine social media platforms automatically. And Remotion generates promotional video clips from the same content, entirely through code.

The total cost? Under $200/month for all four tools combined. The output? More content, more consistently, with better keyword targeting than seven months of agency work ever delivered. Within six weeks, one of my test sites hit 16,000 organic monthly visits and roughly $1,900 in estimated traffic value — from content I never manually wrote, formatted, or published.

I'm going to walk you through the exact setup, every integration point, and the specific decisions that made this work. But first, you need to understand the mistake that makes most people fail at content automation before they even start.

The Strategy Gap That Kills 90% of Content Automation

Here's what happens when most people hear "automate SEO content": they connect an AI writer to a publishing tool, feed it keywords from a spreadsheet, and wait for traffic that never comes.

I know because I did exactly this in early 2025.

I had a keyword list from Ahrefs. High volume terms. Relevant to my niche. I pointed an AI writer at the list and generated 30 posts in a weekend. The writing was technically clean — good grammar, proper structure, decent formatting. And Google ignored every single one of them.

The problem wasn't execution. The problem was strategy.

Every keyword I'd picked looked great in isolation. Solid search volume. Clear commercial intent. But the sites already ranking for those terms had domain authorities three to five times mine. I was generating content for battles I mathematically couldn't win, and no amount of AI writing quality was going to change that equation.

This is the gap that separates automated content that generates actual organic traffic from automated content that generates nothing but server hosting bills. The AI needs to do the thinking before it does the writing. Not "find keywords" — find keywords where there's a measurable gap between what searchers want and what currently ranks. Bottom-funnel terms where the existing top results are thin, outdated, or mismatched to the actual search intent.

When I rebuilt my pipeline with Claude Code at the center, this strategic layer became the foundation everything else rests on. And it changed the entire trajectory of results.

The architecture has four distinct stages, each handled by a specialized tool. Let me break down what each one does and how they connect — because the connections matter more than any individual tool.

The Four-Stage Architecture: How Everything Fits Together

Think of this as a factory with four specialized machines, each doing what it's best at, with Claude Code acting as the floor manager making sure materials flow correctly between stations.

Stage 1 — Claude Code (Strategy & Analysis): Crawls your sitemap, analyzes competitor content, identifies keyword gaps, generates article plans with prioritized keyword clusters, and creates SERP-optimized outlines with H2s, H3s, and semantic structure. This is where intelligence happens.

Stage 2 — Arvow API (Content Generation & Publishing): Takes the article plans and keyword targets from Claude Code, generates fully formatted blog posts with internal linking, images, meta descriptions, and schema markup, then publishes them directly to your CMS. No copy-pasting. No manual formatting. No backend SEO fiddling.

Stage 3 — Blotato API (Social Distribution): Monitors your site's RSS feed for new posts, automatically generates platform-specific social content for Twitter/X, LinkedIn, Instagram, Facebook, TikTok, Pinterest, YouTube, Threads, and Bluesky. Schedules everything through a single API endpoint.

Stage 4 — Remotion (Video Content): Pulls from the same RSS feed and programmatically generates social media videos — slideshows, explainer clips, caption-driven content — using React components rendered to MP4. No video editor. No timeline dragging. Just code.

The power isn't in any single tool. I've written about using Claude Code for SEO analysis and building Remotion video workflows separately. What changed everything was wiring them into a single automated chain where one trigger cascades through all four stages. Publish once, distribute everywhere, create video content — all from the same source material.

Here's how to set up each stage.

Stage 1: Teaching Claude Code to Think Like Your Best SEO Strategist

This is the foundation. Skip this and everything downstream produces mediocre results. Get this right and the entire pipeline compounds.

You're going to create what Claude calls a "skill" — a specialized instruction set that turns Claude Code into a purpose-built SEO analyst. I covered the basics of Claude skills in my SEO content writer skill guide, but the skill we're building here is more ambitious. It doesn't just write — it thinks strategically first.

Setting Up the Workspace

Open Claude Code in VS Code (I prefer it over Claude Desktop for this work because you get full visibility into file operations and can track exactly what the agent is doing). Create a dedicated project folder:

mkdir seo-automation-pipeline
cd seo-automation-pipeline

Inside this folder, create your skill prompt file. This is the instruction set that transforms Claude from a general-purpose assistant into a focused SEO strategist. Here's the structure I use:

# SEO Content Strategist Skill

## Your Role
You are an SEO content strategist for [your business name]. 
Your job is to analyze, plan, and orchestrate — not just write.

## Step 1: Site Analysis
- Crawl the provided sitemap URL
- Catalog all existing content by topic cluster
- Identify content gaps (topics competitors cover that we don't)
- Map internal linking opportunities

## Step 2: Competitor Analysis
- Analyze these competitor sites: [URLs]
- Identify their top-performing content by estimated traffic
- Find keywords where they rank but we don't
- Note content quality weaknesses (thin content, outdated info, 
  poor search intent match)

## Step 3: Keyword Clustering
- Group identified opportunities into topic clusters
- Prioritize by: traffic potential × ranking difficulty × commercial intent
- For each cluster, generate 5-10 article titles with focus keywords
- Flag "quick win" keywords (low competition, decent volume)

## Step 4: Content Brief Generation
- For each approved article, create a SERP-engineered outline
- Include: H2/H3 structure, target word count, semantic keywords,
  internal linking targets, featured snippet opportunities
- Match content type to search intent (how-to, comparison, 
  listicle, deep dive)

The key difference between this and a basic "write me a blog post" prompt is that Steps 1 through 3 happen before any writing begins. Claude Code analyzes your competitive position first, identifies where you can actually win, then creates content plans targeting those specific gaps.

Running the Analysis

Once your skill is configured, trigger it by pointing Claude at your site:

Analyze my site at [your-domain.com/sitemap.xml] 
against these competitors: [competitor1.com], [competitor2.com]
Generate a prioritized content plan for the next 30 days.

Claude Code will crawl your sitemap, fetch competitor pages, and come back with something that looks remarkably like what a $5,000/month SEO consultant would deliver — a prioritized list of keyword clusters, each with article titles, target keywords, estimated traffic values, and difficulty assessments.

The first time I ran this on one of my test sites, Claude identified 47 keyword opportunities I'd missed entirely. Twelve of them were what I'd call "free money" keywords — terms with 500-2,000 monthly searches where the current top results were either outdated by two years or barely addressed the actual search intent. Those twelve articles became my first batch.

You'll get a chance to review and approve each batch before anything gets written or published. This isn't a "set it and forget it" situation at the strategy level — you're the strategic director, the AI is the research team. Approve the keywords that make sense for your business, reject the ones that don't, and move the approved batch to Stage 2.

But here's where it gets interesting. Instead of having Claude Code write the articles itself, I hand them off to a tool that's purpose-built for the writing and publishing layer. And the difference in output quality surprised me.

Stage 2: Arvow Turns Article Plans Into Published Posts

Arvow is the part of the pipeline that handles the work I used to dread most — taking a content brief and turning it into a fully formatted, SEO-optimized, internally linked blog post that's live on my website without me touching a CMS.

What Arvow Actually Does

Think of Arvow as an AI writing team with a built-in web developer. You give it a keyword and an article title (generated by Claude Code in Stage 1), and it produces:

• A complete long-form article (typically 2,000-4,000 words)
• Optimized meta title and description
• Internal links to your existing content (it crawls your sitemap to find them)
• Relevant images and featured graphics
• Proper heading hierarchy (H2s, H3s structured for both readability and SERP features)
• Schema markup appropriate for the content type
• Call-to-action sections matching your brand guidelines

Then it publishes directly to your CMS. If you're on WordPress, Arvow has a plugin that handles the connection. The article goes live, gets auto-submitted to Google Search Console for indexing, and you get a notification.

Connecting Arvow to Claude Code

The integration works through Arvow's API. After Claude Code generates your approved content plan, you can trigger Arvow programmatically:

import requests

ARVOW_API_KEY = "your-api-key"
ARVOW_ENDPOINT = "https://api.arvow.ai/v1/articles"

def publish_article(title, keyword, sitemap_url):
    payload = {
        "title": title,
        "focus_keyword": keyword,
        "sitemap_url": sitemap_url,
        "auto_publish": True,
        "internal_linking": True,
        "generate_images": True,
        "language": "en",
        "tone": "professional",  # matches your brand config
        "word_count_target": 3000
    }
    
    response = requests.post(
        ARVOW_ENDPOINT,
        headers={"Authorization": f"Bearer {ARVOW_API_KEY}"},
        json=payload
    )
    return response.json()

Customizing for Brand Consistency

The part that initially worried me about Arvow was voice. Would the content sound generic? The answer depends entirely on how much configuration work you put in upfront.

Arvow's template system lets you define:

• Brand voice parameters — formal vs. casual, technical depth, sentence structure preferences
• CTA templates — what your calls-to-action say and where they appear
• Internal linking rules — which pages to prioritize for link distribution
• Media preferences — image style, alt text formatting, embedded video handling
• Content structure patterns — whether you prefer listicles, narrative deep dives, or comparison formats

I spent about 40 minutes configuring these settings when I first connected. Since then, every article Arvow produces matches the general brand voice closely enough that a reader wouldn't notice the difference between AI-generated and hand-written posts. The content isn't going to win any creative writing awards — but for SEO content targeting informational queries, it performs exceptionally well.

Here's an honest assessment: Arvow scores roughly 7.5/10 on output quality for general SEO content, according to recent independent reviews. For highly technical or specialized topics, you'll want to edit the output or use Claude Code directly for the writing. But for the high-volume, keyword-targeted content that makes up 70-80% of most SEO strategies? It's more than sufficient, and the automation saves hours per article.

The real magic happens after publishing. Because every article Arvow publishes goes to your site's RSS feed — and that RSS feed is what triggers the next two stages automatically.

Stage 3: Blotato Distributes Every Post Across Nine Platforms

Publishing an article and hoping people find it through Google alone is like opening a restaurant on a side street and hoping foot traffic pays the rent. You need distribution. And doing distribution manually — writing a LinkedIn post, crafting a tweet, formatting an Instagram caption, scheduling a Facebook update — is exactly the kind of repetitive, pattern-based work that should be automated.

Blotato handles this. And it does it through a single API that covers nine platforms simultaneously.

How Blotato Works With Your Content Pipeline

Blotato monitors your website's RSS feed. Every time a new article appears (published by Arvow in Stage 2), Blotato automatically:

1. Pulls the article title, excerpt, and featured image from the RSS entry
2. Generates platform-specific content for each connected account — because what works on LinkedIn doesn't work on TikTok, and a good Instagram caption is nothing like a tweet
3. Creates branded infographics or visual assets for platforms that reward images
4. Schedules posts according to optimal timing for each platform
5. Publishes everything without you touching a button

The platform was built by Sabrina Ramonov, who grew from zero to over a million followers in 13 months using many of these same automation principles. The tool reflects that practical experience — it's not theoretical social media advice packaged as software. It's battle-tested distribution infrastructure.

Setting Up the Integration

You'll need a Blotato account and API access. The setup is straightforward:

import requests

BLOTATO_API_KEY = "your-api-key"
BLOTATO_ENDPOINT = "https://api.blotato.com/v1/posts"

def create_social_posts(article_url, article_title, excerpt):
    payload = {
        "source_url": article_url,
        "title": article_title,
        "excerpt": excerpt,
        "platforms": [
            "twitter", "linkedin", "instagram", 
            "facebook", "tiktok", "pinterest",
            "youtube", "threads", "bluesky"
        ],
        "auto_schedule": True,
        "generate_images": True,
        "brand_template": "default"  # your configured template
    }
    
    response = requests.post(
        BLOTATO_ENDPOINT,
        headers={"Authorization": f"Bearer {BLOTATO_API_KEY}"},
        json=payload
    )
    return response.json()

You can also skip the custom code entirely and use Blotato's native RSS monitoring — just point it at your site's feed URL and it handles detection and posting automatically. Blotato has native integrations with n8n and Make.com, plus REST API support for Zapier and Airtable if you want to add conditional logic.

Platform-Specific Content Generation

This is the piece that separates Blotato from basic cross-posting tools. It doesn't just copy-paste the same text everywhere. Each platform gets content tailored to its format and audience behavior:

• Twitter/X: Short, punchy hooks with a link. Optimized for engagement and retweets.
• LinkedIn: Professional framing with a story-driven opening. Longer format. "Link in comments" convention optional.
• Instagram: Visual-first with branded infographic/carousel. Caption optimized for hashtag discovery.
• Facebook: Community-oriented framing. Question-driven to encourage comments.
• TikTok: Script-ready caption with trending hashtag suggestions.
• Pinterest: Vertical pin graphic with keyword-rich description for Pinterest's search algorithm.
• YouTube: Community post format or Shorts description.
• Threads: Conversational, casual tone. Thread-ready format.
• Bluesky: Adapted for the platform's character limits and link preview behavior.

Blotato supports scheduling up to 900 posts monthly, which gives you enormous headroom even at aggressive publishing frequencies. At 3-4 articles per week across nine platforms, you're looking at roughly 140-160 social posts monthly — well within limits.

If you'd rather have someone build this entire pipeline from scratch, I take on automation and integration projects. You can see what I've built at fiverr.com/s/EgxYmWD.

Stage 4: Remotion Turns Blog Posts Into Videos — Automatically

This is the stage that gets the biggest reaction when I show people the full pipeline. Not because video is inherently more impressive than social posts, but because most people still associate video creation with hours in a timeline editor.

Remotion changes that equation completely. I wrote a detailed guide on creating videos with Claude Code and Remotion and covered the promotional video workflow separately, but here's how it fits into the automated pipeline.

Remotion in 30 Seconds

Remotion is a React framework that treats videos as code. Every frame is a render. Every animation is a function of time. Every visual element is a reusable component. You write (or generate) React components, Remotion renders them as MP4 files.

The integration with Claude Code happened in January 2026 when Remotion launched Agent Skills — 28+ modular instruction files optimized for large language models. The Remotion skill package quickly became the most popular on the skills.sh marketplace, hitting over 25,000 installs within days.

How Video Automation Fits the Pipeline

The video stage monitors the same RSS feed as Blotato. When a new article publishes:

1. Claude Code reads the article content from the RSS feed
2. It extracts key points, statistics, and quotable insights
3. It generates a Remotion composition — React components defining slides, transitions, text animations, and timing
4. Remotion renders the composition to an MP4 file
5. The video is ready for upload to YouTube Shorts, TikTok, Instagram Reels, or any platform that rewards short-form video

The installation is one command:

npx skills add remotion dev skills

Then point Claude at your RSS feed:

Read the latest article from [your-site.com/feed] 
and create a 60-second promotional video summarizing 
the key points. Use vertical aspect ratio for social. 
Add captions.

Claude generates the Remotion code, renders the video, and outputs an MP4. The whole process takes a few minutes — compared to the hours you'd spend cutting, timing, and formatting in a traditional editor.

Customization Options

You're not locked into a single style. The Remotion skill supports:

• Aspect ratios: 16:9 (YouTube), 9:16 (TikTok/Reels/Shorts), 1:1 (Instagram feed)
• Background music: Add licensed audio tracks to your compositions
• Brand colors and fonts: Define once in your Remotion config, apply everywhere
• Animation styles: Slide transitions, fade-ins, kinetic typography, zoom effects
• Caption overlay: Automatically generated and timed to the visual flow

The output quality depends heavily on how much you invest in your template components. A basic setup produces clean, readable social videos that perform well for content distribution. A more polished setup — with custom branded transitions, logo animations, and carefully designed layouts — produces content that looks genuinely professional.

The Desktop vs. VS Code Decision (And Why It Matters More Than You Think)

One decision you need to make early: are you running Claude Code through the Claude Desktop app or through the VS Code extension?

Both work for this pipeline. But they serve different workflows.

Claude Desktop gives you a clean chat interface. You type prompts, Claude responds, files get created. It's simpler and friendlier. The co-work feature lets you run background tasks while you continue chatting. For someone who wants to manage the pipeline conversationally — approving keyword batches, requesting article generation, triggering social posts — Desktop is fine.

Claude Code in VS Code gives you something Desktop doesn't: full visibility into what's happening. You can see the file structure Claude is working with. You can watch it create, modify, and organize files in real time. You can inspect the code it generates before it runs. For developers or power users who want to customize the pipeline, debug integration issues, or extend the skill with new capabilities, VS Code is the clear winner.

My recommendation: start with VS Code. The visibility alone will save you hours of troubleshooting when (not if) something in the pipeline behaves unexpectedly. You can always switch to Desktop later once the pipeline is stable and you're mostly just monitoring.

Where This Approach Honestly Falls Short

I'd be lying if I told you this pipeline is perfect. Seven months of running it — plus extensive testing on multiple sites — has taught me exactly where the cracks are.

Content quality has a ceiling. Arvow produces competent SEO content. It does not produce exceptional content. For your cornerstone articles, thought leadership pieces, or anything targeting competitive head terms — you still need human expertise. I use Arvow for the 70-80% of content that targets mid-tail and long-tail informational queries. The high-stakes 20-30% I either write myself or use Claude Code directly with much more detailed prompting.

Platform algorithm shifts break scheduling logic. Blotato's optimal timing features work great until a platform changes its algorithm. Instagram's engagement patterns shifted noticeably in early 2026, and my scheduled posts were hitting at suboptimal times for about two weeks before I noticed and adjusted. Automation doesn't mean "never check."

Video quality varies wildly. Remotion produces videos as good as the templates and prompts you feed it. A vague prompt generates a vague video. The first few videos I generated looked like corporate slideshow presentations from 2009. It took me roughly a week of template refinement to get output I was genuinely proud of posting.

The strategy layer still needs a human. Claude Code's competitive analysis is impressive, but it can surface keyword opportunities that don't make sense for your specific business context. I've caught it recommending content that was technically in my niche but would attract the wrong audience entirely — people who'd never buy what I sell. You need to review every content plan before it triggers the pipeline.

Costs are real, if lower than an agency. Arvow, Blotato, Claude Code's usage fees, and Remotion's rendering costs add up. My total monthly spend sits around $150-200, which is dramatically less than the $4,200 agency invoice, but it's not free. And if you're running the pipeline across multiple sites, costs scale linearly.

These aren't dealbreakers. They're realities. Anyone selling you fully hands-off content automation without acknowledging these trade-offs is selling you a fantasy. This pipeline reduces manual work by 85-90%. It does not reduce it to zero.

The Results After Six Weeks

Here's what the pipeline produced across two test sites over its first six weeks of operation.

The primary test site went from roughly 800 organic monthly visits to 16,000. The estimated traffic value — what I'd pay for equivalent PPC clicks — reached approximately $1,900/month. That's from a site that had been essentially dormant for eight months before I pointed the pipeline at it.

The content velocity was the biggest shift. Before automation, I was producing maybe one article per week per site if I was disciplined about it. The pipeline publishes 3-4 articles per week, each with social distribution across nine platforms and a companion video clip. That's a 12-16x increase in content output with roughly 2 hours per week of my actual time — mostly spent reviewing and approving Claude Code's content plans.

Social media followers grew steadily but not explosively. The real value of the social distribution wasn't follower counts — it was referral traffic and brand visibility. Posts that linked back to the blog articles drove an additional 15-20% of total traffic on top of organic search.

The video content performed better than I expected on TikTok and YouTube Shorts. Short-form video clips summarizing blog articles consistently hit 2,000-5,000 views, which for a niche tech topic isn't viral but drives a reliable stream of interested visitors back to the full articles.

Pattern I noticed: articles that Claude Code flagged as "quick win" keyword opportunities — low competition, decent volume, poor existing results — were ranking on page one or two of Google within 10-14 days. The articles targeting more competitive terms took 4-6 weeks to start climbing. This matches what most SEO practitioners observe, but having the AI specifically identify and prioritize quick wins meant I saw meaningful traffic growth within the first two weeks rather than waiting months for the more competitive terms to mature.

What I'd Do Differently Starting From Scratch

If I were building this pipeline again today, knowing what I know now, three things would change.

I'd invest more time in Arvow's brand configuration upfront. I rushed through the template setup initially and spent the next two weeks correcting inconsistencies in the published content — wrong CTA placement, internal links pointing to irrelevant pages, meta descriptions that were technically fine but didn't match my brand voice. An extra hour of careful configuration on day one would have saved five hours of corrections over the following weeks.

I'd start with a single platform for Blotato distribution and expand gradually. Launching across all nine platforms simultaneously meant I couldn't tell which platforms were actually driving value. When I eventually dug into the analytics, LinkedIn and Twitter were delivering 80% of the social referral traffic. Pinterest and Threads were essentially zero. I'd have gotten the same results faster by starting with the two highest-impact platforms and adding others after validating their contribution.

I'd build custom Remotion templates before generating any videos. My first batch of auto-generated videos used Remotion's defaults. They were functional but visually bland. After I created branded templates with my color scheme, custom fonts, and animation presets, the video quality jumped dramatically. Front-loading that design work means every video from day one looks polished.

The Bigger Picture: What This Means for Solo Creators and Small Teams

Six months ago, the SEO and content marketing playing field was tilted dramatically toward companies that could afford agencies or dedicated teams. Publishing three articles per week, distributing across nine social platforms, and producing companion videos — that's a three-person team at minimum. A content writer, a social media manager, and a video editor. Call it $8,000-$12,000/month in labor costs for a US-based team. More if you hire an agency.

Today, a solo founder with Claude Code, Arvow, Blotato, and Remotion can match that output for under $200/month and a few hours per week of strategic oversight.

That's not a minor efficiency gain. That's a structural shift in who can compete for organic traffic and social media presence. The barriers to entry for professional-grade content marketing have collapsed, and they're not coming back up.

The catch — and there's always a catch — is that this amplifies your strategy, not just your output. Automate bad strategy and you get bad results faster. The pipeline is only as good as the strategic direction you feed into Stage 1. Claude Code can analyze competitors and find keyword gaps, but only a human who understands their business, their audience, and their positioning can decide which opportunities to pursue and which to ignore.

That's the real skill now. Not writing. Not publishing. Not scheduling social posts. The skill is knowing what to say yes to and what to say no to. Everything else is infrastructure — and infrastructure can be automated.

The tools are here. The integrations work. The question isn't whether automation can handle your content pipeline. The question is whether your strategy is sharp enough to deserve amplification.

Start with Stage 1. Let Claude Code show you the keyword opportunities you've been missing. Approve the ones that fit. Then let the pipeline do what pipelines do — run.

Let's Work Together

Looking to build AI systems, automate workflows, or scale your tech infrastructure? I'd love to help.

• Fiverr (custom builds & integrations): fiverr.com/s/EgxYmWD
• Portfolio: mejba.me
• Ramlit Limited (enterprise solutions): ramlit.com
• ColorPark (design & branding): colorpark.io
• xCyberSecurity (security services): xcybersecurity.io