MCP in Apply – O’Reilly

The next was initially printed in Asimov’s Addendum, September 11, 2025.

Be taught extra concerning the AI Disclosures Mission right here.

1. The Rise and Rise of MCP

Anthropic’s Mannequin Context Protocol (MCP) was launched in November 2024 as a method to make instruments and platforms model-agnostic. MCP works by defining servers and purchasers. MCP servers are native or distant finish factors the place instruments and assets are outlined. For instance, GitHub launched an MCP server that permits LLMs to each learn from and write to GitHub. MCP purchasers are the connection from an AI software to MCP servers—they permit an LLM to work together with context and instruments from completely different servers. An instance of an MCP consumer is Claude Desktop, which permits the Claude fashions to work together with 1000’s of MCP servers.

In a comparatively brief time, MCP has grow to be the spine of a whole bunch of AI pipelines and functions. Main gamers like Anthropic and OpenAI have constructed it into their merchandise. Developer instruments corresponding to Cursor (a coding-focused textual content editor or IDE) and productiveness apps like Raycast additionally use MCP. Moreover, 1000’s of builders use it to combine AI fashions and entry exterior instruments and information with out having to construct a complete ecosystem from scratch.

In earlier work printed with AI Frontiers, we argued that MCP can act as an incredible unbundler of “context”—the information that helps AI functions present extra related solutions to customers. In doing so, it may possibly assist decentralize AI markets. We argued that, for MCP to really obtain its targets, it requires help from:

1. Open APIs: In order that MCP functions can entry third-party instruments for agentic use (write actions) and context (learn)
2. Fluid reminiscence: Interoperable LLM reminiscence requirements, accessed by way of MCP-like open protocols, in order that the reminiscence context accrued at OpenAI and different main builders doesn’t get caught there, stopping downstream innovation

We increase upon these two factors in a latest coverage word, for these trying to dig deeper.

Extra typically, we argue that protocols, like MCP, are literally foundational “guidelines of the highway” for AI markets, whereby open disclosure and communication requirements are constructed into the community itself, somewhat than imposed after the very fact by regulators. Protocols are essentially market-shaping units, architecting markets by the permissions, guidelines, and interoperability of the community itself. They will have a huge impact on how the industrial markets constructed on high of them perform too.

1.1 However how is the MCP ecosystem evolving?

But we don’t have a transparent concept of the form of the MCP ecosystem at present. What are the most typical use instances of MCP? What kind of entry is being given by MCP servers and utilized by MCP purchasers? Is the information accessed by way of MCP “read-only” for context, or does it permit brokers to “write” and work together with it—for instance, by modifying recordsdata or sending emails?

To start answering these questions, we have a look at the instruments and context which AI brokers use by way of MCP servers. This offers us a clue about what’s being constructed and what’s getting consideration. On this article, we don’t analyze MCP purchasers—the functions that use MCP servers. We as a substitute restrict our evaluation to what MCP servers are making accessible for constructing.

We assembled a big dataset of MCP servers (n = 2,874), scraped from Pulse.¹ We then enriched it with GitHub star-count information on every server. On GitHub, stars are just like Fb “likes,” and builders use them to indicate appreciation, bookmark initiatives, or point out utilization.

In observe, whereas there have been loads of MCP servers, we discovered that the highest few garnered many of the consideration and, seemingly by extension, many of the use. Simply the highest 10 servers had practically half of all GitHub stars given to MCP servers.

A few of our takeaways are:

1. MCP utilization seems to be pretty concentrated. Which means, if left unchecked, a small variety of servers and (by extension) APIs might have outsize management over the MCP ecosystem being created.
2. MCP use (instruments and information being accessed) is dominated by simply three classes: Database & Search (RAG), Pc & Internet Automation, and Software program Engineering. Collectively, they obtained practically three-quarters (72.6%) of all stars on GitHub (which we proxy for utilization).
3. Most MCP servers help each learn (entry context) and write (change context) operations, displaying that builders need their brokers to have the ability to act on context, not simply eat it.

2. Findings

To start out with, we analyzed the MCP ecosystem for focus danger.

2.1 MCP server use is concentrated

We discovered that MCP utilization is concentrated amongst a number of key MCP servers, judged by the variety of GitHub stars every repo obtained.

Regardless of there being 1000’s of MCP servers, the highest 10 servers make up practically half (45.7%) of all GitHub stars given to MCP servers (pie chart beneath) and the highest 10% of servers make up 88.3% of all GitHub stars (not proven).

Which means nearly all of real-world MCP customers are seemingly counting on the identical few companies made accessible by way of a handful of APIs. This focus seemingly stems from community results and sensible utility: All builders gravitate towards servers that remedy common issues like net shopping, database entry, and integration with broadly used platforms like GitHub, Figma, and Blender. This focus sample appears typical of developer-tool ecosystems. Just a few well-executed, broadly relevant options are inclined to dominate. In the meantime, extra specialised instruments occupy smaller niches.

2.2 The highest 10 MCP servers actually matter

Subsequent, the highest 10 MCP servers are proven within the desk beneath, together with their star depend and what they do.

Among the many high 10 MCP servers, GitHub, Repomix, Context7, and Framelink are constructed to help with software program growth: Context7 and Repomix by gathering context, GitHub by permitting brokers to work together with initiatives, and Framelink by passing on the design specs from Figma on to the mannequin. The Blender server permits brokers to create 3D fashions of something, utilizing the favored open supply Blender software. Lastly, Activepieces and MindsDB join the agent to a number of APIs with one standardized interface: in MindsDB’s case, primarily to learn information from databases, and in Activepieces to automate companies.

The dominance of agentic shopping, within the type of Browser Use (61,000 stars) and Playwright MCP (18,425 stars), stands out. This displays the elemental want for AI methods to work together with net content material. These instruments permit AI to navigate web sites, click on buttons, fill out types, and extract information similar to a human would. Agentic shopping has surged, regardless that it’s far much less token-efficient than calling an API. Shopping brokers usually have to wade by a number of pages of boilerplate to extract slivers of knowledge a single API request might return. As a result of many companies lack usable APIs or tightly gate them, browser-based brokers are sometimes the only—generally the one—method to combine, underscoring the boundaries of at present’s APIs.

A number of the high servers are unofficial. Each the Framelink and Blender MCP are servers that work together with only a single software, however they’re each “unofficial” merchandise. Which means they don’t seem to be formally endorsed by the builders of the applying they’re integrating with—those that personal the underlying service or API (e.g., GitHub, Slack, Google). As an alternative, they’re constructed by unbiased builders who create a bridge between an AI consumer and a service—usually by reverse-engineering APIs, wrapping unofficial SDKs, or utilizing browser automation to imitate consumer interactions.

It’s wholesome that third-party builders can construct their very own MCP servers, since this openness encourages innovation. But it surely additionally introduces an middleman layer between the consumer and the API, which brings dangers round belief, verification, and even potential abuse. With open supply native servers, the code is clear and may be vetted. Against this, distant third-party servers are tougher to audit, since customers should belief code they’ll’t simply examine.

At a deeper degree, the repos that presently dominate MCP servers spotlight three encouraging info concerning the MCP ecosystem:

1. First, a number of distinguished MCP servers help a number of third-party companies for his or her performance. MindsDB and Activepieces function gateways to a number of (usually competing) service suppliers by a single server. MindsDB permits builders to question completely different databases like PostgreSQL, MongoDB, and MySQL by a single interface, whereas Taskmaster permits the agent to delegate duties to a variety of AI fashions from OpenAI, Anthropic, and Google, all with out altering servers.
2. Second, agentic shopping MCP servers are getting used to get round probably restrictive APIs. As famous above, Browser Use and Playwright entry web companies by an online browser, serving to to bypass API restrictions, however they as a substitute run up in opposition to anti-bot protections. This circumvents the restrictions that APIs can impose on what builders are in a position to construct.
3. Third, some MCP servers do their processing on the developer’s pc (regionally), making them much less depending on a vendor sustaining API entry. Some MCP servers examined right here can run totally on an area pc with out sending information to the cloud—which means that no gatekeeper has the facility to chop you off. Of the ten MCP servers examined above, solely Framelink, Context7, and GitHub depend on only a single cloud-only API dependency that may’t be run regionally end-to-end in your machine. Blender and Repomix are utterly open supply and don’t require any web entry to work, whereas MindsDB, Browser Use, and Activepieces have native open supply implementations.

2.3 The three classes that dominate MCP use

Subsequent, we grouped MCP servers into completely different classes primarily based on their performance.

Once we analyzed what forms of servers are hottest, we discovered that three dominated: Pc & Internet Automation (24.8%), Software program Engineering (24.7%), and Database & Search (23.1%).

Widespread use of Software program Engineering (24.7%) MCP servers aligns with Anthropic’s financial index, which discovered that an outsize portion of AI interactions have been associated to software program growth.

The recognition of each Pc & Internet Automation (24.8%) and Database & Search (23.1%) additionally is sensible. Earlier than the arrival of MCP, net scraping and database search have been extremely built-in functions throughout platforms like ChatGPT, Perplexity, and Gemini. With MCP, nevertheless, customers can now entry that very same search performance and join their brokers to any database with minimal effort. In different phrases, MCP’s unbundling impact is extremely seen right here.

2.4 Brokers work together with their environments

Lastly, we analyzed the capabilities of those servers: Are they permitting AI functions simply to entry information and instruments (learn), or as a substitute do agentic operations with them (write)?

Throughout all however two of the MCP server classes checked out, the preferred MCP servers supported each studying (entry context) and writing (agentic) operations—proven in turquoise. The prevalence of servers with mixed learn and write entry means that brokers usually are not being constructed simply to reply questions primarily based on information but in addition to take motion and work together with companies on a consumer’s behalf.

The 2 exceptions are Database & Search (RAG) and Finance MCP servers, during which read-only entry is a typical permission given. That is seemingly as a result of information integrity is essential to making sure reliability.

3. The Significance of A number of Entry Factors

Just a few implications of our evaluation may be drawn out at this preliminary stage.

First, concentrated MCP server use compounds the dangers of API entry being restricted. As we mentioned in “Protocols and Energy,” MCP stays constrained by “what a specific service (corresponding to GitHub or Slack) occurs to reveal by its API.” Just a few highly effective digital service suppliers have the facility to close down entry to their servers.

One necessary hedge in opposition to API gatekeeping is that most of the high servers strive to not depend on a single presentr. As well as, the next two safeguards are related:

• They provide native processing of knowledge on a consumer’s machine at any time when doable, as a substitute of sending the information for processing to a third-party server. Native processing ensures that performance can’t be restricted.
• If working a service regionally is just not doable (e.g., e-mail or net search), the server ought to nonetheless help a number of avenues of getting on the wanted context by competing APIs. For instance, MindsDB features as a gateway to a number of information sources, so as a substitute of counting on only one database to learn and write information, it goes to nice lengths to help a number of databases in a single unified interface, primarily making the backend instruments interchangeable.

Second, our evaluation factors to the truth that present restrictive API entry insurance policies usually are not sustainable. Internet scraping and bots, accessed by way of MCP servers, are in all probability getting used (at the very least partly) to avoid overly restrictive API entry, complicating the more and more frequent observe of banning bots. Even OpenAI is coloring exterior the API traces, utilizing a third-party service to entry Google Search’s outcomes by net scraping, thereby circumventing its restrictive API.

Increasing structured API entry in a significant approach is important. This ensures that reliable AI automation runs by secure, documented finish factors. In any other case, builders resort to brittle browser automation the place privateness and authorization haven’t been correctly addressed. Regulatory steerage might push the market on this route, as with open banking within the US.

Lastly, encouraging larger transparency and disclosure might assist determine the place the bottlenecks within the MCP ecosystem are.

• Builders working common MCP servers (above a sure utilization threshold) or offering APIs utilized by high servers ought to report utilization statistics, entry denials, and rate-limiting insurance policies. This information would assist regulators determine rising bottlenecks earlier than they grow to be entrenched. GitHub would possibly facilitate this by encouraging these disclosures, for instance.
• Moreover, MCP servers above sure utilization thresholds ought to clearly record their dependencies on exterior APIs and what fallback choices exist if the first APIs grow to be unavailable. This isn’t solely useful in figuring out the market construction, but in addition important info for safety and robustness for downstream functions.

The aim is to not remove all focus within the community however to make sure that the MCP ecosystem stays contestable, with a number of viable paths for innovation and consumer selection. By addressing each technical structure and market dynamics, these instructed tweaks might assist MCP obtain its potential as a democratizing power in AI growth, somewhat than merely shifting bottlenecks from one layer to a different.

Footnotes

1. For this evaluation, we categorized every repo into one among 15 classes utilizing GPT-5 mini. We then human-reviewed and edited the highest 50 servers that make up round 70% of the full star depend in our dataset.

Appendix

Dataset

The total dataset, together with descriptions of the classes, may be discovered right here (constructed by Sruly Rosenblat):

https://huggingface.co/datasets/sruly/MCP-In-Apply

Limitations

There are a number of limitations to our preliminary analysis:

• GitHub stars aren’t a measure of obtain counts and even essentially a repo’s reputation.
• Solely the title and outline have been used when categorizing repos with the LLM.
• Categorization was topic to each human and AI errors and lots of servers would seemingly match into a number of classes.
• We solely used the Pulse record for our dataset; different lists had completely different servers (e.g., Browser Use isn’t on mcpmarket.com).
• We excluded some repos from our evaluation, corresponding to those who had a number of servers and people we weren’t in a position to fetch the star depend for. We could miss some common servers by doing this.

MCP Server Use Over Time