MCP Revenue: How to Price Your Content for AI Agents in 2026

The Model Context Protocol has won. Downloads of the MCP software development kit reached 97 million per month by March 2026, up from 100,000 at launch. 78% of enterprise AI teams already have MCP-backed agents in production. OpenAI, Google, Microsoft, and Salesforce all shipped support within 13 months.

The MCP revenue question is not settled.

Most public MCP servers charge nothing. Publishers who have done the technical work to make their content queryable by AI agents are often not earning a dollar from that access. The infrastructure exists. The AI buyers are there. The missing piece is pricing.

This post covers what to charge, which pricing model fits your content, and what you'll earn at current market rates.

Publishers monetizing content via MCP in 2026 have four pricing models available. Per-query pricing charges AI agents a fee each time they retrieve content, typically in the range of cents per 1,000 queries. Subscription pricing charges a flat monthly or annual fee for a defined access tier. Tiered pricing combines both: a subscription floor with per-query charges above a usage threshold. Outcome-based pricing takes a share of the revenue the AI agent generates using your content.

Per-query is the easiest to deploy and the most common model today. Subscription pricing is preferred by AI platforms that need predictable costs. Tiered pricing is emerging as the enterprise default. Outcome-based is the most sophisticated option, requiring attribution infrastructure most publishers don't yet have.

The right choice depends on one question: how often will AI agents query your content? For news, research, and reference content, per-query fits. For data sources queried thousands of times a day, subscription pricing is more practical for both sides.

Cloudflare Pay-per-Crawl, launched on July 1, 2026, charges AI crawlers between $0.001 and $0.025 per page request. High-traffic sites with tens of millions of monthly page views can earn $50,000 to $200,000 per month. Mid-sized trade publications earn $500 to $5,000. Those are raw crawl rates: AI bots hitting your public website pages and receiving raw HTML.

TollBit, a data marketplace that serves AI companies directly, lets publishers keep 100% of their set rate, with AI buyers paying a transaction fee on top. Mid-market publishers with specialized or premium content typically earn more on TollBit than on Cloudflare Pay-per-Crawl alone, because they can price for content quality rather than traffic volume.

Managed MCP platforms typically take a 20 to 25% cut of revenues they route. That take rate is similar to Cashmere's, which charges roughly 25% on publisher content revenue. If you have 10,000 queries per month at $0.01 each, a 25% platform fee costs $25/month. At 10 million queries, it costs $25,000.

Self-hosted MCP infrastructure lets you keep 100% of every query fee, at the cost of the engineering overhead to run and maintain it. That trade-off is real, and the right answer depends on your query volume and technical resources.

Structured MCP queries and raw web crawls are not the same thing. Pricing them the same way leaves money on the table.

A Cloudflare Pay-per-Crawl request returns raw HTML. The AI receives a copy of your page, parses it, and processes it offline. You have no visibility into what happened next. An MCP query is different: the AI sends a specific question, your endpoint returns a clean, structured, rights-cleared answer, and the AI uses it in real time without retaining your underlying content.

That is a higher-value service. The right benchmark is not "what does a crawl cost" but "what does a structured data API call cost." Financial data APIs, legal research platforms, and scientific abstract databases charge $0.01 to $0.50 per query or more for structured, premium content. That is the correct reference class for publishers pricing MCP access to specialized archives.

Publishers who set their MCP rates at Cloudflare crawl levels are pricing the wrong thing. A structured query to a rights-cleared, authenticated endpoint delivers 10 to 50x more value than a raw crawl. Price accordingly.

The pricing model that earns the most depends on your content's usage pattern.

Per-query pricing fits content with unpredictable, event-driven demand. News archives, scientific papers, legal documents, reference works, and encyclopedic content are queried when an AI agent needs a specific answer to a specific question. The agent queries once, gets the answer, and doesn't return until it needs something else. Per-query pricing captures the full value of each interaction without requiring AI platforms to commit to ongoing access they may not use.

Subscription pricing fits content queried continuously. Financial data, product catalogs, sports statistics, real-time news feeds, and domain-specific databases are queried many times per day by AI systems that depend on them as a standing data source. Those AI platforms want predictable costs. A subscription tier — even a modest one — gives them stable access and gives you stable revenue.

Tiered pricing does both. A subscription floor gives AI platforms baseline access and simplifies their budgeting. Per-query overage pricing ensures that high-usage months contribute proportionally to your revenue. This is the model behind Alien Intelligence's IaaS structure: a fixed infrastructure fee for baseline access, and a per-usage layer that captures the upside as query volume grows.

Publishers who run their own MCP infrastructure retain full pricing power. Those who use managed platforms trade some of that control for reduced technical overhead. That trade-off is often less obvious than it appears.

Managed platforms enforce their own pricing floors, take-rate caps, and payment routing. You can't easily experiment with custom pricing tiers, bundle MCP access with a subscription product, or negotiate direct commercial terms with a specific AI buyer — those transactions go through the platform. And if the platform's terms change, your economics change with them.

Self-hosted MCP infrastructure, the model Alien Intelligence's data streaming platform uses, deploys directly on your servers. You control the pricing logic, the access rules, and the billing integration. Every dollar stays with you.

This matters especially for publishers with EU data residency requirements. Most managed MCP platforms are US-hosted. Routing EU publisher content through US-hosted infrastructure may conflict with GDPR requirements and AI Act compliance obligations. Self-hosted, on-premises deployment is often the only option for national archives, academic publishers, and regulated broadcasters in Europe. The same choice that satisfies data sovereignty requirements also gives you full pricing control.

Setting a price is one thing. Enforcing it is another.

The IAB Tech Lab's Content Monetization Protocol, finalized in April 2026, is designed to solve exactly this. CoMP requires AI systems to establish commercial agreements with publishers before crawling or using their content. It supports per-crawl, aggregation, and outcome-based models, and provides a standardized mechanism for attaching pricing terms to the content access channel itself.

Publishers who embed CoMP-compliant terms in their MCP endpoints change the dynamic from "I hope you pay me" to "you can't access this without agreeing to pay." AI agents that discover your endpoint through a CoMP-compliant discovery mechanism see the pricing terms before the first query.

This is the infrastructure model Alien Intelligence is built around: pricing is enforced at the data access layer, not the legal layer. An AI system that doesn't pay doesn't get content. That's a cleaner enforcement mechanism than any contract, and it's what makes the per-query pricing model viable at scale.

MCP is already the default protocol for AI agent data access. The infrastructure is there. The buyers are there. What most publishers are missing is the pricing layer that turns access into revenue.

The per-query model works best for on-demand content like news, research, and reference archives. Subscription pricing fits continuously queried data sources. Tiered pricing captures both. Outcome-based pricing is the ambition for publishers who can prove their content's contribution to AI outputs.

Structured MCP access should command higher rates than raw crawls, because it delivers more value: a clean, rights-cleared answer rather than raw HTML. The correct reference point is premium structured data APIs, not crawl rates.

And your deployment choice determines how much of that revenue you keep. Managed platforms reduce engineering overhead but take a cut and limit pricing flexibility. Self-hosted infrastructure gives full pricing control and is the only compliant option for EU publishers with data residency requirements.

Price before you deploy. Don't discover the revenue model after.

Explore Alien Intelligence's data streaming infrastructure for publisher pricing control or see what publishers are already earning from AI licensing in 2026.

Revenue varies widely by content type, query volume, and pricing model. As a reference point, Cloudflare Pay-per-Crawl rates of $0.001 to $0.025 per page request allow high-traffic sites to earn $50,000 to $200,000 per month, while mid-sized trade publications earn $500 to $5,000. Structured MCP queries, which return clean rights-cleared answers rather than raw HTML, should command higher rates. Premium structured data APIs for financial, legal, and scientific content typically charge $0.01 to $0.50 per query. Publishers pricing their MCP endpoints in that range rather than at crawl rates capture significantly more revenue per AI interaction.

Pay-per-Crawl charges AI bots when they fetch raw pages from your public website. The AI receives a raw HTML copy and processes it offline. Per-query MCP pricing charges AI agents when they send a structured query to your content endpoint and receive a structured, rights-cleared answer in real time. MCP queries deliver more value to the AI buyer and give the publisher better attribution and logging. Publishers should price MCP access at a premium over Pay-per-Crawl rates for this reason.

Managed platforms reduce technical overhead but take a 20 to 25% revenue share and limit pricing flexibility. Self-hosted MCP infrastructure requires more engineering investment but gives full pricing control and keeps 100% of revenue. For publishers in the EU, self-hosted is often the only option that meets GDPR and AI Act data residency requirements, since most managed platforms are US-hosted. Publishers with high query volumes and the engineering capacity to run their own infrastructure typically earn more from self-hosted deployment over time.

CoMP (Content Monetization Protocol), finalized by the IAB Tech Lab in April 2026, is a standard that requires AI systems to establish commercial agreements with publishers before accessing their content. It supports per-crawl, aggregation, and outcome-based pricing models, and provides a mechanism to embed pricing terms directly in the content access protocol. Publishers using CoMP-compliant infrastructure change the dynamic from hoping AI companies pay to requiring payment as a condition of access. AI systems that discover your content through a CoMP-compliant endpoint see the pricing terms before their first query.

EU publishers face additional constraints from GDPR and AI Act data residency requirements that limit which deployment models they can use. Most managed MCP platforms are US-hosted, which means routing EU publisher content through them may not meet data residency requirements. This pushes EU publishers toward self-hosted infrastructure deployed on their own servers. That constraint is also a pricing advantage: self-hosted infrastructure gives full pricing control that managed platforms restrict. EU publishers who build compliant data infrastructure automatically gain the pricing flexibility that publishers on managed platforms lack.