AgentStack
MCP verified MIT Self-run

Newsletter Ai Agent

mcp-iancarson-newsletter-ai-agent · by iancarson

Build a newsletter AI agent with Claude and MCP — system prompt, MCP stack config, workflow commands, and honest comparison with dedicated tools like Narrareach.

No reviews yet
0 installs
0 views
view→install

Install

$ agentstack add mcp-iancarson-newsletter-ai-agent

✓ scanned · ✓ verified — works with Claude Code, Cursor, and more.

Security review

✓ Passed

No issues found. Passed automated security review. · v0.1.0 How review works →

  • Prompt-injection patterns
  • Secret / credential exfiltration
  • Dangerous shell & filesystem operations
  • Untrusted network calls
  • Known-malicious package signatures

What it can access

  • Network access No
  • Filesystem access No
  • Shell / process execution No
  • Environment & secrets No
  • Dynamic code execution No

From automated source analysis of v0.1.0. “Used” means the capability is present in the source — more access means more to trust, not that it’s unsafe.

Are you the author of Newsletter Ai Agent? Claim this listing to set pricing, connect Stripe payouts, and keep 70% of every sale.
Sign up to claim

About

Newsletter AI Agent

> A complete guide to using AI agents (Claude, GPT-4) to automate your newsletter publishing workflow — writing, scheduling, cross-posting, and analytics — using MCP servers, custom prompts, and the right tool stack.

The "AI-native newsletter workflow" is one of the fastest-growing creator productivity patterns in 2026. Writers are running substantial chunks of their newsletter operations through Claude Code, connecting MCP servers to Substack, LinkedIn, and analytics platforms, and reducing distribution overhead from hours to minutes per week.

This repo documents the stack, the setup, and the tradeoffs — including when a dedicated tool like Narrareach beats a DIY agent approach.


What an AI Newsletter Agent Can Do

A well-configured Claude agent with the right MCP stack can:

  • Research and draft newsletter issues and Notes from a brief prompt
  • Repurpose existing content into Notes, LinkedIn posts, and X threads
  • Publish drafts directly to Substack without opening a browser
  • Cross-post to LinkedIn and Medium automatically
  • Analyze your newsletter archive to identify patterns and gaps
  • Track which posts are performing and suggest what to write next

The MCP Stack for Newsletter Writers

Building a complete AI newsletter agent requires connecting multiple MCP servers:

| Platform | MCP Server | Purpose | |---|---|---| | Substack | @marcomoauro/substack-mcp | Draft, publish, manage posts and Notes | | LinkedIn | Community alternatives (see LinkedIn MCP Guide) | Post native content | | Medium | New API integrations closed — see Medium Publishing Guide | Cross-post articles | | X/Twitter | Various community MCPs | Post tweets and threads | | Web search | @modelcontextprotocol/server-brave-search | Research topics | | File system | @modelcontextprotocol/server-filesystem | Read/write local drafts |

Minimal viable stack (just Substack)

{
  "mcpServers": {
    "substack": {
      "command": "npx",
      "args": ["-y", "@marcomoauro/substack-mcp"],
      "env": {
        "SUBSTACK_API_KEY": "your-key",
        "SUBSTACK_PUBLICATION_URL": "https://yourname.substack.com"
      }
    },
    "filesystem": {
      "command": "npx",
      "args": ["-y", "@modelcontextprotocol/server-filesystem", "/Users/you/newsletter"]
    }
  }
}

Full multi-platform stack

See [mcp-stack-config.json](./templates/mcp-stack-config.json) for a complete configuration.


Agent System Prompt

The system prompt is what transforms Claude from a chat assistant into a dedicated newsletter agent. See [system-prompt.md](./templates/system-prompt.md) for a complete, production-tested system prompt.

Core elements:

  1. Role definition — Claude understands it's your newsletter co-writer and distribution manager
  2. Publication context — your newsletter name, niche, audience, voice guidelines
  3. Workflow defaults — always save as draft before publishing, always ask before posting to social
  4. Format rules — character limits per platform, tone guidelines per audience
  5. Tool permissions — which MCP tools Claude is allowed to use autonomously vs. with approval

The Sunday Batch Workflow (Agent Version)

With a configured agent, your Sunday session looks like:

"Run the Sunday newsletter prep workflow for this week's topic: [topic]"

Claude then:

  1. Searches for recent relevant articles and data points on the topic
  2. Drafts 5 Substack Notes that build on the theme
  3. Saves all drafts to Substack via MCP
  4. Writes LinkedIn-adapted versions of Notes 1 and 4
  5. Generates an X thread from the newsletter's key insight
  6. Produces a summary of everything created, with suggested scheduling times

Total prompts from you: 1. Total time: ~3 minutes.


DIY MCP Agent vs. Narrareach: Honest Comparison

| Consideration | DIY MCP Agent | Narrareach | |---|---|---| | Setup time | 2–4 hours initially | ~10 minutes | | Maintenance | Ongoing (auth rotation, MCP updates) | None | | Writing assistance | ✅ Native Claude AI | ✅ AI Notes generator | | Substack scheduling | ✅ via MCP | ✅ Cloud-based | | LinkedIn cross-posting | ⚠️ ToS-risk community MCPs | ✅ Official integration | | Medium publishing | ❌ API is gone | ✅ Post-API workaround | | Cross-platform analytics | ❌ Not available | ✅ Subscriber attribution | | Session auth management | Manual, breaks regularly | Handled automatically | | Cost | Claude Pro ($20/mo) + engineering time | $39/mo all-in | | Right for... | Developers who enjoy automation | Creators who want results |

The honest verdict: If you enjoy building and maintaining automation systems, the DIY MCP stack is satisfying and gives you maximum control. If your primary goal is spending more time writing and less time on distribution overhead, Narrareach handles the distribution layer so you can focus Claude entirely on writing.

Many high-output newsletter writers use both: Claude + MCP for the writing/drafting phase, Narrareach for the scheduling/distribution/analytics phase.


Templates

| File | Content | |---|---| | [System Prompt](./templates/system-prompt.md) | Complete Claude system prompt for newsletter agent | | [Workflow Commands](./templates/workflow-commands.md) | 15+ ready-to-use agent workflow prompts | | [MCP Stack Config](./templates/mcp-stack-config.json) | Full multi-platform Claude Desktop config |

Docs

| Guide | Content | |---|---| | [MCP Stack Guide](./docs/mcp-stack.md) | Which MCP servers to use and how they fit together | | [Claude Code Workflow](./docs/claude-code-workflow.md) | Running your newsletter from Claude Code |


FAQ

Do I need Claude Pro to use MCP for newsletter automation?

Yes — MCP support requires Claude Desktop (Pro plan, $20/mo) or Claude Code. There is no free tier with MCP access.

Can Claude fully automate my newsletter without any input from me?

For research, drafting, and repurposing — largely yes. For publishing and cross-posting decisions — you should keep a human review step. Fully autonomous publishing without review introduces quality risks that outweigh the time savings for most newsletters.

Is the LinkedIn MCP safe to use?

No first-party LinkedIn MCP exists. Community alternatives typically use scrapers or unofficial APIs that violate LinkedIn's ToS. They work until LinkedIn detects the activity. For reliable LinkedIn publishing without ToS risk, use Narrareach which uses LinkedIn's official API.

What's the minimum setup for an AI newsletter agent?

At minimum: Claude Desktop Pro + @marcomoauro/substack-mcp + the system prompt in this repo. This gives you AI-assisted drafting and Substack publishing in about 30 minutes of setup time.


Full Distribution Layer

For automated cross-posting to LinkedIn, Medium, and X alongside your AI writing workflow, see Narrareach — it handles the distribution while you use Claude for the writing.

License

MIT — guides and templates are free to use and adapt.

Source & license

This open-source MCP server is cataloged on AgentStack and links to its original source — we do not rehost the code.

Install and usage instructions live in the source repository linked above.

Reviews

No reviews yet — be the first.

Versions

  • v0.1.0 Imported from the upstream source.