MCP in Apply – O’Reilly

The next was originally published in Asimov’s Addendum, September 11, 2025.

Study extra in regards to the AI Disclosures Venture here.

1. The Rise and Rise of MCP

Anthropic’s Model Context Protocol (MCP) was launched in November 2024 as a solution 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 enables LLMs to each learn from and write to GitHub. MCP purchasers are the connection from an AI utility 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 hundreds of MCP servers.

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

In earlier work printed with AI Frontiers, we argued that MCP can act as an important unbundler of “context”—the information that helps AI functions present extra related solutions to shoppers. In doing so, it might probably assist decentralize AI markets. We argued that, for MCP to really obtain its objectives, it requires assist 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 through 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 recent policy note, for these seeking to dig deeper.

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

1.1 However how is the MCP ecosystem evolving?

But we don’t have a transparent thought of the form of the MCP ecosystem immediately. What are the commonest use circumstances of MCP? What kind of entry is being given by MCP servers and utilized by MCP purchasers? Is the information accessed through 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 through 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 out there 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 knowledge on every server. On GitHub, stars are much like Fb “likes,” and developers use them to indicate appreciation, bookmark initiatives, or point out utilization.

In apply, whereas there have been loads of MCP servers, we discovered that the highest few garnered many of the consideration and, doubtless 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. Because of this, if left unchecked, a small variety of servers and (by extension) APIs may have outsize management over the MCP ecosystem being created.
2. MCP use (instruments and knowledge being accessed) is dominated by simply three classes: Database & Search (RAG), Pc & Net 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 assist 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 begin with, we analyzed the MCP ecosystem for focus threat.

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 hundreds 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).

Because of this the vast majority of real-world MCP customers are doubtless counting on the identical few companies made out there through a handful of APIs. This focus doubtless stems from community results and sensible utility: All builders gravitate towards servers that resolve common issues like internet shopping, database entry, and integration with broadly used platforms like GitHub, Figma, and Blender. This focus sample appears typical of developer-tool ecosystems. A number of 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 improvement: 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 utility. Lastly, Activepieces and MindsDB join the agent to a number of APIs with one standardized interface: in MindsDB’s case, primarily to learn knowledge 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 techniques to work together with internet content material. These instruments permit AI to navigate web sites, click on buttons, fill out types, and extract knowledge similar to a human would. Agentic shopping has surged, regardless that it’s far much less token-efficient than calling an API. Looking brokers usually have to wade via a number of pages of boilerplate to extract slivers of information a single API request may return. As a result of many companies lack usable APIs or tightly gate them, browser-based brokers are sometimes the best—typically the one—solution to combine, underscoring the bounds of immediately’s APIs.

Among the high servers are unofficial. Each the Framelink and Blender MCP are servers that work together with only a single utility, however they’re each “unofficial” merchandise. Because of this they don’t seem to be formally endorsed by the builders of the appliance they’re integrating with—those that personal the underlying service or API (e.g., GitHub, Slack, Google). As a substitute, they’re constructed by impartial 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. Nevertheless it 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. In contrast, distant third-party servers are tougher to audit, since customers should belief code they’ll’t simply examine.

At a deeper stage, the repos that at the moment dominate MCP servers spotlight three encouraging info in regards to the MCP ecosystem:

1. First, a number of outstanding MCP servers assist a number of third-party companies for his or her performance. MindsDB and Activepieces function gateways to a number of (usually competing) service suppliers via a single server. MindsDB permits builders to question completely different databases like PostgreSQL, MongoDB, and MySQL via 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 doubtlessly restrictive APIs. As famous above, Browser Use and Playwright entry web companies via an internet 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 capable of construct.
3. Third, some MCP servers do their processing on the developer’s laptop (domestically), making them much less depending on a vendor sustaining API entry. Some MCP servers examined right here can run solely on an area laptop with out sending knowledge to the cloud—that means that no gatekeeper has the ability 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 domestically end-to-end in your machine. Blender and Repomix are fully 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 & Net 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 economic index, which discovered that an outsize portion of AI interactions have been associated to software program improvement.

The recognition of each Pc & Net Automation (24.8%) and Database & Search (23.1%) additionally is sensible. Earlier than the arrival of MCP, internet 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 very 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 knowledge 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 most well-liked 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 will not be being constructed simply to reply questions primarily based on knowledge but additionally to take motion and work together with companies on a consumer’s behalf.

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

3. The Significance of A number of Entry Factors

A number of 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 Power,” MCP stays constrained by “what a specific service (equivalent to GitHub or Slack) occurs to reveal via its API.” A number of highly effective digital service suppliers have the ability to close down entry to their servers.

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

• They provide native processing of information on a consumer’s machine every time 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 operating a service domestically shouldn’t be doable (e.g., electronic mail or internet search), the server ought to nonetheless assist a number of avenues of getting on the wanted context via competing APIs. For instance, MindsDB capabilities as a gateway to a number of knowledge sources, so as a substitute of counting on only one database to learn and write knowledge, it goes to nice lengths to assist a number of databases in a single unified interface, basically making the backend instruments interchangeable.

Second, our evaluation factors to the truth that present restrictive API entry insurance policies will not be sustainable. Net scraping and bots, accessed through MCP servers, are in all probability getting used (no less than partly) to bypass overly restrictive API entry, complicating the increasingly common apply of banning bots. Even OpenAI is coloring outdoors the API strains, utilizing a third-party service to entry Google Search’s outcomes via internet scraping, thereby circumventing its restrictive API.

Increasing structured API entry in a significant means is significant. This ensures that official AI automation runs via 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 could push the market on this path, as with open banking within the US.

Lastly, encouraging better transparency and disclosure may assist establish the place the bottlenecks within the MCP ecosystem are.

• Builders working widespread 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 knowledge would assist regulators establish rising bottlenecks earlier than they grow to be entrenched. GitHub may facilitate this by encouraging these disclosures, for instance.
• Moreover, MCP servers above sure utilization thresholds ought to clearly listing 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 additionally important info for safety and robustness for downstream functions.

The objective 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 may assist MCP obtain its potential as a democratizing power in AI improvement, reasonably than merely shifting bottlenecks from one layer to a different.

Footnotes

1. For this evaluation, we categorized every repo into one in every of 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 complete dataset, together with descriptions of the classes, may be discovered right here (constructed by Sruly Rosenblat):

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

Limitations

There are just a few limitations to our preliminary analysis:

• GitHub stars aren’t a measure of obtain counts and even essentially a repo’s recognition.
• Solely the identify 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 doubtless match into a number of classes.
• We solely used the Pulse listing 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, equivalent to people who had a number of servers and people we weren’t capable of fetch the star depend for. We could miss some widespread servers by doing this.

MCP Server Use Over Time