Key takeaways
- Free tier drives discovery and adoption; pro tier drives revenue: the free tier is your marketing funnel
- Split by depth (free = real-time data, pro = analysis and screening) not by basic functionality
- Use separate repos: public for free (builds trust), private for pro (protects source)
Why freemium works for MCP servers
People do not pay for tools they have never tried. They pay for tools they already use and want more of.
The freemium model splits your MCP server into two tiers:
- Free tier: core tools, no license key, public repo. Drives discovery and adoption.
- Pro tier: advanced tools, license key required, private repo. Drives revenue.
Users find your free server, install it, start relying on it, and then upgrade when they need the advanced features. The free tier is not lost revenue: it is your marketing funnel.
Case study: FinAgent
FinAgent is a real MCP server on the marketplace that uses this model. Here is how it is split:
Free tier: finagent-mcp
| Tool | What it does |
|---|---|
get_stock_data | Current and historical stock prices |
get_market_news | Financial news for any ticker |
get_company_info | Company profile and key stats |
Package: finagent-mcp on PyPI Repo: Public on GitHub: anyone can inspect the source code
No license key. No restrictions. Install it, use it, done.
Pro tier: finagent-pro-mcp
Everything in the free tier, plus:
| Tool | What it does |
|---|---|
screen_stocks | Screen stocks by financial criteria |
get_sec_filings | SEC filing search and analysis |
get_insider_trades | Insider trading activity |
get_earnings_calendar | Upcoming earnings dates |
Package: finagent-pro-mcp on PyPI Repo: Private on GitHub: source code not visible to buyers
Requires a license key (MCP_LICENSE_KEY) purchased on MCP Marketplace.
Why this works
- Free tier builds trust. Users see exactly what they are getting: the code is open source, the tools work, the data is real.
- Free tools are genuinely useful. Stock prices and market news cover the most common use cases. Users do not feel like they are getting a crippled demo.
- Pro tools are a natural upgrade. Once you are using FinAgent for stock data, stock screening and SEC filings are the obvious next step.
- Two marketplace listings. The free listing appears in search results and category pages. It links to the pro listing in its description. Double the surface area.
How to structure your free/pro split
Deciding what is free vs pro
The split should feel natural, not arbitrary. Rules of thumb:
Make it free if:
- It is the core use case people search for
- It uses free or public APIs
- It establishes the value of your server
- Competitors offer it for free
Make it pro if:
- It requires expensive API calls or proprietary data
- It solves a deeper problem that power users care about
- It is the "next step" after the free tools prove value
- It is hard to replicate from public sources
Bad split: Free gives you read access, pro gives you write access. This feels arbitrary: users expect basic CRUD.
Good split: Free gives you real-time data, pro gives you historical analysis and screening. This feels like a natural depth progression.
The repos
- Free tier: Public GitHub repo. Users can inspect the source, file issues, and see your code quality. This builds trust.
- Pro tier: Private GitHub repo. Source code is not visible. The package is still on PyPI as a public listing: users install it with
pip install, but they cannot read the source.
Both repos publish to PyPI independently. Different package names, different versions, independent release cycles.
Cross-promotion
Your free tier's README and marketplace listing should mention the pro tier:
"Need stock screening and SEC filings? Upgrade to FinAgent Pro."
One line. No hard sell. Just make the upgrade path visible.
Implementation
You create two separate packages: a free server and a pro server. The free server has no license logic. The pro server includes all tools but gates the premium ones with a license key check.
With MCP Creator, scaffold each separately: add paid=true and paid_tools to the pro server. Available for TypeScript (mcp-creator-typescript) and Python (mcp-creator-python).
Manually, use the mcp-marketplace-license SDK to add license checks to pro tools. See the license key guide for code examples and the monetization guide for the complete setup.
Pricing your pro tier
One-time vs subscription
- One-time ($5-25): Simple, low-friction purchase. Good for utilities where the value is static.
- Monthly ($5-25/mo): Same local install model, but recurring. Best when you ship ongoing updates and new features — buyers expect continued value.
- Subscription ($10-50+/mo): Good for remote hosted servers, data feeds, API-dependent tools, or servers with ongoing infrastructure costs.
Most MCP servers start with one-time pricing. You can always add a subscription tier later.
Finding the right price
- Check comparable tools on MCP Marketplace for reference
- Start at the low end: you can always raise prices once you have traction
- The free tier already handles discovery, so pro pricing can reflect real value without worrying about scaring away first-time users
Marketplace commission
MCP Marketplace takes a commission on sales. See the integration page in your creator dashboard for current rates and payout details.
Common mistakes
Making the free tier useless. If free tools are too limited to be genuinely useful, nobody installs your server and the funnel breaks. The free tier should solve a real problem on its own.
Too many tiers. Free and pro is enough. Adding "basic," "standard," "premium," and "enterprise" tiers for an MCP server is overkill. Two tiers. Keep it simple.
Same repo for both. Use separate repos. The pro repo should be private so buyers cannot just read the source. The free repo should be public to build trust.
Not linking them together. If your free and pro listings do not mention each other, you lose the upgrade funnel. Cross-link in READMEs, marketplace descriptions, and blog posts.
Next steps
- How to monetize your MCP server: license keys, remote hosting, and the full monetization guide
- License keys for MCP servers: how the verification SDK works
- How to publish your MCP server to PyPI: packaging and distribution
- MCP Creator: scaffold free and pro servers with AI in TypeScript or Python
- How to build an MCP server: the full build guide, start here if you haven't built your server yet