← All publications
5000 Emails Turned into 896 LinkedIn Posts
I turned 5,000 emails into 896 LinkedIn content briefs.
Without reading a single email myself. Here's the full technical setup.
The problem:
5 years of marketing newsletters buried in Gmail. Quality content from experts, hidden under thousands of emails.
Read them one by one? Impossible. Delete them? Wasteful.
The solution: Claude Code + MCP
MCP (Model Context Protocol) is an open standard by Anthropic that lets AI connect to external tools. My AI can read Gmail and write to Obsidian directly.
```
Gmail (MCP Server)
↓ read emails
Claude Code (orchestrator)
↓ analyze + generate
Obsidian (MCP Server)
↓ write briefs
Vault "Knowledge Base"
```
Step 1: Gmail connection
A Gmail MCP server exposes search, read, and label functions.
Gmail caps at 500 results per query, so I paginated with `before:YYYY/MM/DD` to scan chronologically across 5 years.
Step 2: Smart filtering
From 5,000+ emails, I built a filtering pipeline:
• Identified high-quality educational senders
• Subject deduplication
• Excluded pure promos, webinars, discount codes
Result: ~50% contained real actionable content.
Step 3: Parallel processing
Instead of 1 email at a time, I launched up to 15 AI agents in parallel. Each received 20-40 emails:
• Read the full email via Gmail
• Create a structured summary
• Generate 3-5 TIMELESS LinkedIn post ideas
• Save to Obsidian in a standardized format
Step 4: Standardized format
Every brief follows the same template: YAML metadata + structured summary + 3-5 evergreen post ideas. Tags enable instant retrieval by topic.
Results:
• 896 briefs in Obsidian
• ~3,000 LinkedIn post ideas
• 5 years covered (2021-2026)
• ~30 agents across 3 sessions
• Estimated cost: ~$100-150 in tokens
Key takeaways:
1. Your emails are a sleeping asset. Experts spent hours writing quality content that sits unread in your inbox.
2. MCP is a game changer. Connecting AI to your existing tools unlocks possibilities we couldn't imagine 6 months ago.
3. Parallelization is everything. One agent = hours. Thirty agents = minutes per batch.
4. The "timeless" filter matters. No dates, no events. Only content that holds up over time.
Next: turning those 3,000 ideas into an automated editorial calendar.
Questions about the setup? Drop them in the comments.
Without reading a single email myself. Here's the full technical setup.
The problem:
5 years of marketing newsletters buried in Gmail. Quality content from experts, hidden under thousands of emails.
Read them one by one? Impossible. Delete them? Wasteful.
The solution: Claude Code + MCP
MCP (Model Context Protocol) is an open standard by Anthropic that lets AI connect to external tools. My AI can read Gmail and write to Obsidian directly.
```
Gmail (MCP Server)
↓ read emails
Claude Code (orchestrator)
↓ analyze + generate
Obsidian (MCP Server)
↓ write briefs
Vault "Knowledge Base"
```
Step 1: Gmail connection
A Gmail MCP server exposes search, read, and label functions.
Gmail caps at 500 results per query, so I paginated with `before:YYYY/MM/DD` to scan chronologically across 5 years.
Step 2: Smart filtering
From 5,000+ emails, I built a filtering pipeline:
• Identified high-quality educational senders
• Subject deduplication
• Excluded pure promos, webinars, discount codes
Result: ~50% contained real actionable content.
Step 3: Parallel processing
Instead of 1 email at a time, I launched up to 15 AI agents in parallel. Each received 20-40 emails:
• Read the full email via Gmail
• Create a structured summary
• Generate 3-5 TIMELESS LinkedIn post ideas
• Save to Obsidian in a standardized format
Step 4: Standardized format
Every brief follows the same template: YAML metadata + structured summary + 3-5 evergreen post ideas. Tags enable instant retrieval by topic.
Results:
• 896 briefs in Obsidian
• ~3,000 LinkedIn post ideas
• 5 years covered (2021-2026)
• ~30 agents across 3 sessions
• Estimated cost: ~$100-150 in tokens
Key takeaways:
1. Your emails are a sleeping asset. Experts spent hours writing quality content that sits unread in your inbox.
2. MCP is a game changer. Connecting AI to your existing tools unlocks possibilities we couldn't imagine 6 months ago.
3. Parallelization is everything. One agent = hours. Thirty agents = minutes per batch.
4. The "timeless" filter matters. No dates, no events. Only content that holds up over time.
Next: turning those 3,000 ideas into an automated editorial calendar.
Questions about the setup? Drop them in the comments.