MCP-Billing Review (2026): Pricing, Features & Honest Verdict

MCP-Billing Review: The Verdict

MCP-Billing is a self-hosted boilerplate that gives an MCP server OAuth 2.1, API key management and Stripe usage-based billing already wired together. You buy the source once for EUR 79 and own it, with no revenue share. Best for: developers monetising an MCP server. Price: EUR 79 one-time, rising to EUR 149. Rating: 7/10.

We found MCP-Billing at #10 on Product Hunt with 114 upvotes. It is not a SaaS, it is a codebase, and knowing that up front changes how you should evaluate it.

What is MCP-Billing?

Anyone who has tried to charge money for an MCP server has hit the same wall. The protocol part is easy and well documented. The commercial part is not: you need OAuth 2.1 with PKCE so clients can authenticate, API keys so machines can, metering that survives retries and half-finished streams, rate limiting so one client cannot bankrupt you, and Stripe wired to all of it. Each piece is solved in isolation. Nobody ships them working together, and the integration is where the weeks disappear.

MCP-Billing is that integration, sold as source code. Built by Marc Gil, an individual developer rather than a company, it is a Next.js 15 project spanning seven modules with more than 300 passing tests across 29 test files. You pay once, download it, and it is yours for unlimited personal and client projects under a commercial licence. There is a 7-day refund guarantee.

The detail that suggests the author has actually run this in production is how the metering handles edge cases. Every usage event carries an idempotency key, so replaying the same key bills once rather than twice. And one usage event is recorded per completed stream rather than per token, which is the correct call: token-level metering on a streamed response that dies halfway leaves you either overcharging a customer for output they never received or eating the cost yourself. Those two decisions are the difference between billing code that works and billing code that generates support tickets.

What Are MCP-Billing’s Key Features?

OAuth 2.1 with PKCE, already implemented

Full OAuth 2.1 plus PKCE without you reading the RFCs. This is the module that saves the most calendar time, because OAuth is not conceptually hard so much as unforgiving, and the gap between an implementation that works in testing and one that is correct is where people lose a fortnight. MCP clients increasingly expect standards-compliant OAuth, so this is not optional if you want to be listed anywhere.

Stripe usage-based billing with the edge cases handled

Charging per call or per token, with explicit handling for retries, partial failures and streamed responses. The idempotency key on every usage event means a network retry does not double-bill, and metering per completed stream avoids the partial-response problem. If you have written billing code before you know these are the exact three places it breaks, and they are usually discovered in production by an annoyed customer rather than in testing.

API key management with zero-downtime rotation

A dashboard for generating, rotating and revoking keys, with rotation that does not break live clients. Multi-tenancy is built in, so each client gets an isolated context and its own quota. Rate limiting runs on Redis to stop one caller from eating your capacity or your model budget. These are unglamorous and you will need every one of them the first time a customer’s script goes into a loop.

A free, open-source metering core

The metering middleware is published separately as mcp-metering, an MIT-licensed npm package, free to use in production without buying anything. This is a smart bit of product design and also a genuinely useful hedge. If all you need is idempotent usage metering, take the free package and build the rest yourself. The paid boilerplate is for people who want the auth, key management, rate limiting and Stripe wiring as well.

How Much Does MCP-Billing Cost?

MCP-Billing is EUR 79 as a one-time payment during the launch window, roughly $90, rising to EUR 149 afterwards. There is no subscription, no revenue share and no platform fee. The licence covers unlimited personal and client projects, so an agency can buy it once and use it across every client build, which is where the value gets silly in a good way. Refunds within 7 days. Seven days is short for a codebase you need to actually integrate before you can judge it, so plan to spend the first evening reading the source rather than leaving it in a folder until the weekend.

Compare that to the alternatives on total cost rather than sticker price. A managed auth provider plus a billing platform is typically $25 to $100+/mo before you have a single paying customer, and it scales with your user count forever. Revenue-share platforms take a percentage of everything you ever earn. MCP-Billing costs less than two hours of developer time and then costs nothing again. If it saves the two to three weeks the site claims, and OAuth plus Stripe metering plausibly is two to three weeks, the return is not a close call.

The honest caveat is what one-time pricing means for you, not for the seller. You own the code, which also means you maintain it. When the MCP spec moves, when Stripe deprecates an API version, when Next.js 15 becomes Next.js 17, that is your problem. A SaaS charging monthly is also selling you the promise that someone else handles that. There is no published update policy or maintenance commitment here, and for a one-person project that is the risk you are accepting for the price.

Who is MCP-Billing Best For?

Use MCP-Billing if you have a working MCP server and want to charge for it. That is the exact case it is built for, and buying three weeks back for EUR 79 is an easy decision if your server is already doing something people want. It is a particularly good buy for agencies and consultants, since one licence covers unlimited client projects and the second deployment has effectively paid for the first ten.

Use it if you are comfortable in the Next.js and TypeScript world and want to own your billing stack rather than rent it. Owning the source means you can change the pricing model, add a tier, or move billing providers without asking permission or waiting for a roadmap.

Skip MCP-Billing if your stack is not Next.js. This is not a language-agnostic SDK, it is a Next.js 15 application, and porting the concepts to Python or Go means you are writing it yourself anyway and should just take the free metering package. Skip it if you want a managed service with a support contract and an SLA, because a one-time purchase from an individual developer is not that. And skip it if your MCP server is free, internal, or still a weekend experiment, since there is nothing to bill.

Best MCP-Billing Alternatives

Building it yourself with Clerk or WorkOS plus Stripe is the default path. You get managed auth with real support, and Stripe’s own metering primitives are solid. Expect roughly $25 to $100+/mo in recurring platform costs depending on user count, plus the integration work, which is the part MCP-Billing is actually selling you. Reasonable if you want a vendor to call when auth breaks at 3am.

Stripe’s own MCP tooling covers the payments and metering side and is free beyond normal Stripe fees. What it does not give you is OAuth 2.1, API key management, multi-tenancy or rate limiting, so you are assembling most of the stack regardless. Fine if billing is the only piece you are missing.

The free mcp-metering package from the same author is the alternative worth naming explicitly, because it is the honest answer for a lot of people. MIT licensed, free, production-ready idempotent metering. If you already have auth solved and only need usage tracking that survives retries, take this and keep your EUR 79.

Final Verdict: Is MCP-Billing Worth It?

MCP-Billing is a well-scoped answer to a problem that is about to get much more common. MCP servers are multiplying, a growing number of them are worth money, and almost none of their authors want to spend three weeks on OAuth flows and Stripe webhooks before finding out whether anyone will pay. At EUR 79 with no revenue share, the arithmetic barely needs doing. The idempotency and stream-completion decisions in the metering layer show real production experience, and open-sourcing the metering core is a confident move that makes the paid product easier to trust rather than harder.

What holds it back is what it is rather than how good it is. A boilerplate transfers maintenance to you the moment you buy it, and this one comes from a single developer with no stated update commitment, in a protocol ecosystem that is still changing shape month to month. Next.js 15 only is a hard constraint that rules out a large share of potential buyers. And the price doubling to EUR 149 after launch changes the calculation, though not enough to make it a bad deal. Recommended for Next.js developers with a paid MCP server to ship, with clear eyes about owning the code afterwards. Rated 7/10.