Key takeaways
- Two models: license key + pip package ($5-25 one-time, or monthly for ongoing updates) or remote hosted server ($10-50+/mo)
- The
mcp-marketplace-licenseSDK handles verification with one line of code - Combine free and paid tiers for the most effective pricing strategy
You built something valuable
You built an MCP server that solves a real problem. People are using it. Now you want to get paid for your work.
MCP Marketplace supports two monetization models. Each fits different use cases, and you can use both at once (free tier for discovery, paid tier for power features).
Not sure how to build an MCP server yet? Start with How to build an MCP server first. Once you have something to monetize, come back here.
Two models for paid MCP servers
Model 1: License key + package ($5-25 one-time, or monthly)
Your server ships as a normal pip or npm package. Users install it locally. A license key gates premium tools at runtime. Price one-time for static tools, or monthly if you ship ongoing updates and new features.
How it works:
- User purchases your server on MCP Marketplace
- They receive a license key (
mcp_live_...) - They set
MCP_LICENSE_KEYas an environment variable - Your server verifies the key at startup (local cache for fast startup, graceful fallback)
- Premium tools unlock
Best for: utilities, data tools, integrations, anything where the code runs locally.
Model 2: Remote hosted server ($10-50+/month)
Your server runs on your infrastructure. Users connect to a URL instead of installing a package. They never see your code.
How it works:
- User subscribes on MCP Marketplace
- They add your server URL to their MCP config
- Their AI assistant connects to your server over HTTP
- You handle authentication, rate limiting, and access control
Best for: proprietary data, AI models, anything where you want full access control. Naturally piracy-proof since no code is distributed.
How license verification works
The mcp-marketplace-license SDK handles verification automatically: install it with pip install mcp-marketplace-license, and it reads the license key from environment, caches locally for fast startup, and verifies against the API when needed. See the license key guide for full details.
Building a paid local server
Path A: Use MCP Creator (fastest)
Scaffold a paid server with AI using mcp-creator-typescript (TypeScript) or mcp-creator-python (Python). Just add paid=true:
scaffold_server(
package_name="my-paid-mcp",
description="My paid MCP server",
tools=[...],
paid=true
)
This generates a project with:
- License SDK in dependencies (
@mcp_marketplace/licensefor TypeScript,mcp-marketplace-licensefor Python) - License verification wired into every tool
.env.examplewithMCP_LICENSE_KEY- README with license setup instructions
Want to gate only specific tools? Pass paid_tools:
scaffold_server(
package_name="my-freemium-mcp",
description="Free and pro tools",
tools=[...],
paid=true,
paid_tools=["pro_tool_1", "pro_tool_2"]
)
Path B: Add licensing manually
Install the SDK and add one line to gate your entire server:
from mcp_marketplace_license import with_license
with_license(mcp, slug="my-server")
Or gate individual tools for a freemium model. See the license key guide for both approaches with full code examples.
Building a remote hosted server
Path A: Use MCP Creator
Both mcp-creator-typescript and mcp-creator-python support hosting="remote":
scaffold_server(
package_name="my-remote-mcp",
description="A remote hosted MCP server",
tools=[...],
hosting="remote"
)
This generates:
- Server entry point using HTTP transport (not stdio)
- A
Dockerfilefor easy deployment - README with remote MCP config examples
.env.examplewithPORT=8000
Combine with paid=true for a hosted server that also validates license keys:
scaffold_server(
package_name="my-premium-remote-mcp",
tools=[...],
paid=true,
hosting="remote"
)
Path B: Set up manually
Switch your server's transport from stdio to HTTP:
def main():
port = int(os.environ.get("PORT", "8000"))
mcp.run(transport="sse", host="0.0.0.0", port=port)
Deploy to any cloud provider (Railway, Fly.io, AWS, your own VPS). Users connect with:
{
"mcpServers": {
"your-server": {
"url": "https://your-server.com/mcp"
}
}
}
The free/pro model
The most effective pricing strategy combines free and paid tiers. Free tools drive discovery and adoption. Pro tools generate revenue.
FinAgent is a real example of this pattern:
- Free tier (finagent-mcp): stock data and market news: public repo, no license needed
- Pro tier (finagent-pro-mcp): SEC filings and stock screening: private repo, license key required
Users discover FinAgent through the free tier, try it, and upgrade when they need the advanced tools.
Pricing strategy
One-time vs subscription
- One-time ($5-25): Good for utilities and tools where the value is static. Simple for buyers. You get paid once per user.
- Monthly ($5-25/mo): Same license key model, but recurring. Best when you ship ongoing updates, new features, or expanding tool sets. Buyers expect continued value.
- Subscription ($10-50+/mo): Good for remote hosted servers, data feeds, AI models, and anything with ongoing infrastructure costs. Recurring revenue.
Commission
MCP Marketplace takes a commission on sales. See the creator docs for current rates.
Pricing tips
- Start lower than you think. You can always raise prices.
- Look at comparable tools on the marketplace for reference.
- Free tier is not lost revenue: it is your marketing funnel.
Submit to the marketplace
Once your paid server is built:
- Publish to PyPI (Python) or npm (TypeScript)
- Go to Submit on MCP Marketplace
- Set your pricing type (one-time or subscription) and price
- Your server gets listed with a purchase button
For remote servers, submit the URL instead of a PyPI package name.
Summary
| License key + pip | Remote hosted | |
|---|---|---|
| Price range | $5-25 one-time or /mo | $10-50+/mo |
| User gets | pip/npm package + license key | URL to connect to |
| Code visible | Yes (private repo on GitHub) | No (runs on your server) |
| Piracy risk | Low (key verified via API) | None (code never shared) |
| Best for | Utilities, integrations | Proprietary data, AI models |
| Infra needed | None | Your server/cloud |
Both models work. Pick the one that fits your server, or use both with a free/pro split.
Deep dives
- License keys for MCP servers: how the verification SDK works, testing, and security
- Free vs Pro: the freemium model with a real-world case study
- Remote MCP servers: SSE transport, Docker, and deployment options
- MCP Creator: scaffold paid servers with AI in TypeScript or Python
- How to build an MCP server: the full build guide before you monetize