As GitHub adds MCP support to Copilot, Anthropic updates its protocol and OpenAI adopts Anthropic’s standard, it’s clear that the Model Context Protocol (MCP) is quickly becoming the connective tissue between AI and APIs. But behind the headlines of major tech companies embracing this new standard lies a surprising truth: Most organizations are still struggling with basic API implementation, let alone AI integration.
In a revealing conversation with Sagar Batchu, CEO and co-founder of Speakeasy, we uncovered how his company is helping bridge this gap through automation, while also addressing the emerging challenges of the MCP ecosystem.
The Persistent API Challenge
“The core business problem we are solving at Speakeasy is that unlocking the value of APIs is pretty challenging for a business,” explains Batchu. “Building great APIs, if you’re a tech company, exposing your services to the world is a tough problem.”
It’s surprising that despite being a focus for many companies for years, Batchu estimates only about 10% of Fortune 500 companies are effectively managing their API programs. This represents a significant opportunity for both disruptors and established companies willing to invest in better API infrastructure.
“The reason Stripe exists is for that reason,” Batchu notes, highlighting how upstarts with superior developer experiences can challenge industry giants.
From Developer Experience to Agent Experience
The rapid evolution of AI has expanded the API integration challenge. Now, companies need to serve human developers and AI agents attempting to use their services.
“Your user base has evolved from developers to AI,” says Batchu. “Many of the AI products we use, like ChatGPT and Claude, also use your APIs.”
This is where MCP enters the picture. Created by Anthropic, MCP standardizes how applications provide context to large language models, functioning as Batchu calls “a protocol and a format to what is otherwise just undefined.”
Speakeasy is seeing adoption trends, with developer products leading the charge, followed by data-heavy services like ETL providers and database companies. These early adopters recognize the potential of bringing “the power of LLMs’ unstructured reasoning capabilities into the world of all the data you have.”
The MCP Maturity Challenge
While MCP is gaining momentum, Batchu emphasizes that effective implementation isn’t as simple as exposing an existing API through the protocol.
“There’s a real need for a curation process,” Batchu explains. “A company can’t just take its API and throw it out whole hog into the AI world. It needs to trim it down, decide what goes into it and figure out the right descriptions to expose.”
This challenge is evident in the fact that, according to Batchu, only about 20% of companies that have released MCP servers have ones that work well. The successful ones invest time in curating their interfaces for AI consumption.
Speakeasy helps automate this process, allowing companies to focus on their core business while ensuring their APIs are accessible to traditional developers and AI agents.
The Future of the Protocol Landscape
While MCP has “the wind in its sails,” according to Batchu, the landscape could still evolve. “First to market doesn’t always necessarily win,” he notes. “It’s going to be what gets the right developer integrations and developer buy-in.”
Speakeasy remains protocol-agnostic, ready to support whatever dominant standards emerge. “We are the buffer between the ecosystem and how [companies] build their business,” says Batchu, positioning his company as the translation layer between various protocols and implementation methods.
“The Model Context Protocol (MCP) is filling a crucial gap between AI models and APIs, a necessary step as AI becomes more integrated into software and services. However, the path to realizing MCP’s full potential isn’t uniform, as implementation quality and completeness can differ significantly among vendors,” said Mitch Ashley, VP and Practice Lead DevOps and Application Development at The Futurum Group. “As model and agent APIs and frameworks expand and evolve and become agentic, the complexity of managing and navigating these interfaces will necessitate solutions for curation and developer/agent experience.”
The Path Forward
For companies looking to prepare for this AI-integrated future, Batchu offers clear advice: it all starts with a great API specification. “You want to start from documenting and managing a great specification for your API,” he explains, as this “unlocks an ecosystem of tooling and ability to get live with an MCP or get live with the next protocol very quickly.”
As MCP moves from bleeding edge to lighthouse adopters, Batchu expects it to become standard within 12–18 months. Companies that prepare now by focusing on solid API foundations will be positioned to take advantage of this shift, empowering their human developers and AI agents to unlock the full value of their services.
Read more about how Agentic Agents and MCP Signal Shift to Development Work.
