Agentic

Content

System

Build 03

©2026 Nikhil Khedlekar. All rights reserved.

Core Expertise

Generative AI Workflows • Prompt Engineering • Content Automation • Content Strategy • Programmatic SEO • CMS Architecture • Search Intent Strategy • WordPress • Figma • Canva • Analytics

Back to Tools
Build 03

Agentic AI Content Production System

A Claude Code workflow that researches, plans, drafts, illustrates and SEO-checks an article, then hands a ready WordPress draft to a human editor, so the repetitive production steps stop eating the day.

By Ongoing since Aug 2026 Updated

Self-initiated build · Claude Code · WordPress REST API · Aug 2026 – Present
Agent-Driven Tool Selection
Search Console + Trends Validated
Canva MCP Imagery
Human Editorial Gate
Stack
Claude Code · MCP · WordPress REST API
Role
Workflow design & orchestration
Status
Ongoing build
Timeline
Aug 2026 – Present
In short

An agentic content production system built on Claude Code. It researches, drafts, illustrates and SEO-checks an article, then prepares a WordPress draft for a human editor. It never publishes on its own.

The Daily Grind

The Same Steps, Every Article

The steps are small and the repetition is what hurts. That is the part an agent can carry.

Publishing one article is never one task. It is a keyword check, a trend check, an outline, research, a draft, internal links, external sources, an image, an SEO pass, and then copying everything into WordPress and formatting it.

Each step is small. Repeated for every article, they become most of the working day, and the part that actually needs a person (judgement, voice, accuracy) gets squeezed.

  • The same checklist rebuilt from memory for every post
  • Copy-paste between five or six tools
  • Quality drifting as the day gets longer
  • Editorial time spent on logistics instead of editing

This system moves the logistics to an agent and keeps the editorial judgement with a person. It is the tooling counterpart to the process I redesigned in the AI Editorial Engine case study.

Design Principles

Rules Before Prompts

The system is not a clever prompt. It is an agent that follows written rules, the way a new editor would:

  • Content rules, editorial standards and business voice are written down once and applied every time
  • The agent picks the tool each task needs: research, keyword data, images, linking
  • It prepares a draft, never a published post
  • A human editor reviews everything before it goes live

The rules are the product. The model underneath is replaceable.

System Architecture

Tools the Agent Orchestrates

Each integration removes a copy-paste step. The agent decides which one a task needs.

Orchestrator
Claude Code
Integrations
Model Context Protocol
Visuals
Canva MCP
Search Data
Google Search Console
Trend Validation
Google Trends
Publishing
WordPress REST API
Core Logic

How It Works

A linear production line with a human gate at the end, and a feedback loop that improves the rules.

Brief
A topic or target keyword starts the run.
Research & keyword analysis
The agent researches the topic on the web and analyses the keyword and its search intent.
Content plan
It outlines the article against the content rules, mapping sections to what readers are searching for.
Draft
It writes to the editorial standards and business voice defined in its instructions.
Imagery
Images are generated through the Canva MCP integration.
SEO validation
Google Trends and Search Console data check the topic and keyword choices before anything is finalised.
Linking
Internal links to existing content and external links to credible sources are added.
WordPress draft
The REST API creates a draft post with content, images and metadata in place.
If the editor approves
They publish it.
If it needs changes
They edit it, and the lesson is folded back into the rules.
Guardrails

Guardrails That Stay

Automation is only useful if it can be trusted. These are fixed in every run.

Drafts only, never auto-publish
Human editorial review before publication
Written voice and standards applied in every run
Sources and links checked as part of the workflow
How it was built: The work was designing the system, not writing code: deciding the rules, the tools, the order of steps and where the human sits. Claude Code does the execution inside that structure.
Roadmap

What I'd Improve Next

Feed briefs from Search Console query data on a schedule
Audit internal links across the existing library
Score drafts against the rules before they reach the editor
Log editor changes to refine the rules automatically
Key Learnings

What This Taught Me

  • Agent quality comes from clear rules and the right tools, not from longer prompts.
  • Integrations (MCP, REST APIs) are where the real time savings are, because they remove the copy-paste.
  • A human gate before publishing is a feature, not a limitation.
  • Writing down what good looks like is the hardest and most valuable step.
Capabilities

What This Demonstrates

Agentic workflow design and tool orchestration
MCP and REST API integrations
Translating editorial standards into rules a machine can follow
Search-data-driven content decisions
Human-in-the-loop system design
FAQ

Common Questions

What is the Agentic AI Content Production System?

It is a Claude Code workflow that researches a topic, plans and drafts an article to written editorial rules, generates images through the Canva MCP integration, validates keywords with Google Trends and Search Console data, adds internal and external links, and prepares a WordPress draft through the REST API for a human editor.

Does it publish articles by itself?

No. It creates drafts only. A human editor reviews every piece before it is published.

What are the "rules" the agent follows?

Content rules, editorial standards, business voice and execution requirements, written down once and applied in every run, so quality does not depend on a long prompt.

Which tools does the agent use?

Claude Code orchestrates the work. It uses web research, Google Search Console and Google Trends for keyword and trend validation, the Canva MCP integration for images, and the WordPress REST API for drafts. The agent chooses the tool each step needs.

How does the system improve over time?

When an editor changes a draft, the lesson is folded back into the written rules, so the next draft starts closer to done.

Keep Exploring

Related Work

Case Study AI Editorial Engine Rebuilding a slow, bureaucratic publishing workflow into an AI-assisted pipeline. Read more Case Study Search Authority System Pillar-cluster architecture built around search intent. Read more Build Watchtower A self-healing monitor that emails me the moment something new appears. Read more
Next build
Audio Analysis Automation

Want to automate
your content ops?

Built to carry the repetitive production steps and leave the editorial judgement to a person.