The $0 Distribution Playbook for MCP Servers: 12 Channels, 30 Days, Real Numbers
A month ago I shipped aicraft-code-review — an open-source MCP server that reviews code locally (OWASP scanning, N+1 detection, style rules) for Claude Code, Cursor and Cline. I had $0 marketing budget and no audience. One month later: 650+ PyPI installs, 12 Product Hunt followers, and listings in 5 directories — with more pending. Zero dollars spent. Here's the exact playbook, including what…
In a month, a developer launched an open-source MCP server called aicraft-code-review that reviews code locally for various tools. Despite having no marketing budget and no audience, the server achieved impressive results within that short period. PyPI installations reached over 650, Product Hunt gained 12 followers and generated 7 comments, and it was listed in 5 directories, with more pending. The developer shared their "playbook" detailing the channels used to achieve these results and what didn't work.
The channels were ranked by the effort-to-reach ratio and cost, all of which were $0. PyPI had 650+ installs after an afternoon of setting up the package. Smithery, another directory, approved the server in a day, which helped the cursor.directory submission. Product Hunt gave a chance to gather valuable feedback for future updates.
Glama is still reviewing the server, and the awesome-mcp-servers repository is a significant discovery point for MCP servers. However, some channels like Reddit and Hacker News didn't yield the expected results, so the developer advises building karma in technical subreddits before promoting the project.
Key learnings from the playbook include: shipping on PyPI/npm first, preparing common requirements like Dockerfile, LICENSE, SECURITY.md, glama.json, and install docs, submitting to free directories in parallel, capturing feedback from various sources, and focusing on the two big PRs - awesome-mcp-servers and the Docker MCP Registry - once the server's listing quality score is improved.
Written by urgent.news from Dev.to's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.