The MCP ecosystem is moving fast enough that a useful server can go from "published" to "effectively abandoned" in a week — not because it broke, but because you posted once on launch day and never again. Most indie builders don't have a distribution habit. They have a shipping habit.
I built marketing-pipeline to handle the recurring work I kept skipping: rotating posts across channels, keeping directory listings current, showing up consistently without me thinking about it.
The mechanics: onboarding a project takes one command.
marketing onboard --name my-tool --repo owner/repo --kind mcp-server
That fetches the README, sends it to Claude, and writes problem, facts, and angles to projects.yml. From there, a daily GitHub Actions cron at 14:00 UTC picks the least-recently-used angle per project and drafts posts at per-channel length limits — 300 chars for Bluesky, 280 for X, 150–400 words for Dev.to and Hashnode.
The kind field does the routing. Tag something mcp-server and it targets MCP Registry, Smithery, Glama, and PulseMCP. Tag it claude-skill and it goes toward awesome-claude-code (the pipeline generates the payload; their rules require a human to submit via their GitHub issue form, so that step stays manual). Tag it browser-extension and it routes to Chrome Web Store, Firefox AMO, and Edge Add-ons.
The part I spent the most time on is pipeline/antislop.py — a hard-reject gate that runs before any post is published. It blocks specific tokens: excited, game-changer, unlock, empower, AI-powered, emoji, hashtags, exclamation points, rhetorical questions. Soft filtering wasn't enough; I needed something that would fail loudly rather than let a garbage post through at 2am.
Required credentials: Anthropic API key, plus Bluesky, Dev.to, and Hashnode tokens. Mastodon and Slack are optional.
If you've shipped an MCP server or Claude Code skill and the launch-day post was the last one, this is the part of the stack that was missing.
github.com/robertnowell/marketing-pipeline
Top comments (0)