Content Generation with AI Agents: Blog Posts, Docs & Social Media

Step-by-Step Setup

The following instructions configure the core content generation stack: Notion MCP, Brave Search MCP, Filesystem MCP, and Slack MCP. WordPress Skill is added as the final publishing step.

Step 1: Create API Credentials for Each Service

Before editing your MCP config, collect the necessary credentials:

• Notion: Create an Integration at notion.so/my-integrations and share relevant databases with it
• Brave Search: Sign up at brave.com/search/api and generate a free API key (2,000 queries/month)
• Slack: Create a Slack App with channels:read, chat:write, and files:write scopes
• WordPress: Enable Application Passwords in your WP admin and create one for the agent user

Step 2: Add Servers to Your MCP Config

Open ~/.claude/settings.json (Claude Code) or .cursor/mcp.json (Cursor):

{
  "mcpServers": {
    "notion": {
      "command": "npx",
      "args": ["-y", "@modelcontextprotocol/server-notion"],
      "env": {
        "NOTION_API_KEY": "your_notion_integration_token"
      }
    },
    "brave-search": {
      "command": "npx",
      "args": ["-y", "@modelcontextprotocol/server-brave-search"],
      "env": {
        "BRAVE_API_KEY": "your_brave_api_key"
      }
    },
    "filesystem": {
      "command": "npx",
      "args": ["-y", "@modelcontextprotocol/server-filesystem",
               "/path/to/content-library"]
    },
    "slack": {
      "command": "npx",
      "args": ["-y", "@modelcontextprotocol/server-slack"],
      "env": {
        "SLACK_BOT_TOKEN": "xoxb-your-slack-bot-token"
      }
    },
    "wordpress": {
      "command": "npx",
      "args": ["-y", "mcp-server-wordpress"],
      "env": {
        "WP_URL": "https://yourblog.com",
        "WP_USER": "agent-publisher",
        "WP_APP_PASSWORD": "xxxx xxxx xxxx xxxx xxxx xxxx"
      }
    }
  }
}

Step 3: Store Brand Guidelines Locally

Create a brand-voice.md file in your content library directory. Include your tone of voice description, approved vocabulary, forbidden phrases, target audience persona, and 2–3 example paragraphs that exemplify the desired style. The agent will load this file via Filesystem MCP at the start of every drafting session.

Step 4: Verify Each Connection

Restart your AI assistant and test with lightweight prompts:

• "List the last 5 pages in my Notion content calendar database" — verifies Notion MCP
• "Search Brave for the latest statistics on AI content generation 2026" — verifies Brave Search MCP
• "Read the brand-voice.md file from my content library" — verifies Filesystem MCP
• "Post a test message to the #content-drafts Slack channel" — verifies Slack MCP

Workflow: Research → Draft → Review → Publish → Promote

The complete AI content production workflow spans five phases. A single instruction can trigger all five end-to-end, or you can run each phase interactively with checkpoints.

Phase 1: Research

The agent queries Notion MCP to pull the content brief for the target article — title, target keyword, audience, and any specific angles to cover. It then runs 3–5 Brave Search queries to gather current statistics, identify top-ranking competitor articles, and find expert perspectives. Research findings are stored in a local notes file via Filesystem MCP.

Phase 2: Draft

The agent loads the brand-voice.md guidelines from the filesystem, reviews the research notes, and writes a full draft. For blog posts this includes a headline, meta description, introduction, body sections with H2/H3 structure, internal link suggestions, and a conclusion with a CTA. The draft is saved to the Notion page and to a local Markdown file simultaneously.

Phase 3: Review

The agent posts a Slack message to the designated review channel via Slack MCP: "Draft ready: [Notion page link]. Reply APPROVE to publish or leave feedback." It reads the thread response and applies requested revisions. If the reply contains APPROVE, the workflow advances to publication. If it contains feedback text, the agent revises the draft and posts an updated version for re-review.

Phase 4: Publish

On approval, the agent calls WordPress Skill to create a new post with the draft content, assigns the appropriate category and tags derived from the Notion brief, and either publishes immediately or schedules it for the date specified in the content calendar. The Notion page status field is updated to "Published" via Notion MCP.

Phase 5: Promote

After publishing, the agent generates 3–5 social media variants of the article — a Twitter/X thread, a LinkedIn post, and a short newsletter blurb — and posts them to the appropriate Slack channels for distribution. For social channels with direct MCP integrations, the agent can post directly; otherwise the formatted copy is ready for manual scheduling.

Use Cases

SEO Blog Post Pipeline

Content teams running monthly SEO programs use this stack to produce 20–40 articles per month without proportionally scaling headcount. The agent handles research and first drafts; human editors review and approve. Time per article drops from 4 hours to 45 minutes of human attention.

Technical Documentation Updates

When a new API version ships, the agent reads the changelog via Filesystem MCP, identifies which Notion documentation pages need updating, generates revised content sections, and creates a Slack notification for the docs team to review. Documentation stays current without a dedicated technical writer managing every update.

Product Launch Content Batch

For product launches, the agent generates the full content package in a single session: press release draft, blog announcement, 5-tweet launch thread, LinkedIn post, and email newsletter — all in brand voice, all grounded in the product spec loaded via Filesystem MCP. The package is reviewed in one Slack thread and published across channels on approval.

Multilingual Content Adaptation

For teams publishing in multiple languages, the agent reads an English source article via Filesystem MCP, adapts it for each target locale (adjusting idioms, currency, and examples via Brave Search for local context), and creates separate WordPress posts or Notion pages per language. Filesystem MCP handles per-locale brand voice guidelines so the translated content sounds natural, not machine-translated.

Frequently Asked Questions