The tech industry is poised for a seismic shift as generative AI becomes more autonomous. AI agents are not just traversing documents or external links — they’re beginning to call production APIs, perform complex queries, make mutations, and orchestrate sequential calls. At the heart of this movement is the Model Context Protocol (MCP), an open-source standard protocol that streamlines API integration for AI agents.

Industry experts anticipate that AI agents are about to hammer APIs with unprecedented levels of traffic. As such, many technology providers have begun to wrap their APIs in MCP servers and re-evaluate their digital strategies to prepare for agents. So, where does this all leave developer portals, which are traditionally designed for humans?

Put simply, these new consumption modes don’t replace the need for developer portals — not by a long shot. The rise in AI agents cements both the role of APIs and the discipline of maintaining organized, specification-driven documentation with rich developer-centric resources — areas that modern developer portals particularly excel at. That said, there are steps developer portal owners can take to adapt to machine-driven API consumption.

AI Agents: The New API Consumer

Humans have traditionally interacted with APIs through hardcoded integrations, but AI agents will soon do so autonomously on the fly. Although AI agents only currently account for just 1% of API traffic, Gartner expects that figure to surge to 30% by 2029. Companies are actively turning to AI agents in business areas like sales pipelines, customer support, finance, and other backend tasks.

MCP is playing a key role in standardizing and redefining these integration patterns at scale, providing a means for agents to connect with APIs. We’ve seen this really grow legs in developer tooling, with AI agents such as Anthropic’s Claude, Cursor, OpenAI’s ChatGPT, Google Gemini, or Windsurf, providing support for MCP. Using MCP, developers can link large language models (LLMs) to compliant servers that expose a manifest and describe API functionality in a format agents can understand.

For instance, OpenCage, the popular geocoding API, recently launched an MCP server to support features like forward geocoding and reverse geocoding. WireMock also hosts an MCP server that connects with AI coding agents to help simulate APIs. We’ve seen official ones emerge from AWS, GitHub, Couchbase, and plenty of others. Major AI companies and hyperscalers alike already support MCP, and the number of public MCP servers continues to increase.

Why Developer Portals Matter For AI Agents

So, what does this surging interest in machine-driven API connectivity mean for human-readable documentation and developer portals? Although AI agents may soon consume APIs directly, it doesn’t displace your typical developer resources — it actually heightens their need. More than ever, you’ll require a system of record, a control plane, and a human governance layer for APIs.

Developer portals act as a source of truth for your digital ecosystem. A solid developer portal aggregates disparate APIs and specifications across clouds and organizations and serves them up in a digestible format with a strong developer experience. An enterprise-grade portal also supports API governance, offering an embedded content management system and role-based access to segment information to various API stakeholders.

A developer portal is a great location to extend this rich knowledge for machines. One effective way is to host or point to an MCP server. This is more than just a Google search to find documentation — it’s a tailored and secure context engine to provide other agents with the right (and relevant) information for their task at hand. The task-oriented nature of MCP tools makes them highly complementary to developer portals that house reference docs, workflows, and real-world usage patterns.

Including observability features in your developer portal can produce valuable insight into LLM-based traffic and usage patterns. For instance, it could reveal which specific AI agents are calling which APIs. Having such knowledge could greatly inform your product planning, developer relations, and overall marketing strategy.

Lastly, developer portals are critical to maintain as a check and balance on AI. AI agents are prone to mistakes and hallucinations, failing simple tasks up to 70% of the time, reports The Register. Most research demonstrates that human-in-the-loop workflows tend to produce the best results, too. As such, technical documentation with rich developer resources will always be complementary to agentic consumption, since it provides a means to debug and fix issues generated by AI agents.

Evolving Developer Portals To Guide AI Agents

Developer portals are foundational — they aren’t going anywhere. But this isn’t to say they should remain static in the face of drastic shifts in technology patterns. Instead, there are exciting opportunities for API publishers to double down on their developer centers to better position themselves for agentic consumption. Let’s review some of these strategies.

Thoroughly Document API Mechanics

Whereas humans get bogged down in context, AI requires rich context to properly understand the request and provide a quality response. A developer portal houses additional context that an AI can train on. To set your portal up for success, go beyond OpenAPI specifications to provide further context, such as server timeouts, sample code in various languages, error messages, and resolution steps. Thoroughly documenting API mechanics reduces the potential for hallucination and provides more understanding to complement an MCP tool, which is intended to have a narrow focus.

Describe Common Usage Patterns

Another area where API developer portals shine is showcasing common usage patterns. This can range from user journeys and case studies, in the form of blog posts, to more technical documentation of multi-step API workflows, such as authentication flows, long-running server processes, or chained DevOps-related actions.

API flows could even be described using Arazzo Specification, a standard format for outlining a sequential series of API calls for specific use cases. Adding such context to your portal is important to train AI agents on common patterns they can then recommend to end users. If you’re packaging such actions up in an MCP server, this can act as a proof of record for their behaviors.

Enhance Generative AI Discovery

AI agents can’t use your API if they can’t find it. Thankfully, developer portals can take steps to adapt to the emerging practice of generative engine optimization (GEO). One key strategy is to make machine-readable integration assets easily discoverable, such as:

• Official MCP servers
• Community-supported MCP servers
• Other AI standard formats (A2A, Agent-Protocol, etc.)
• OpenAPI definitions
• Arazzo specifications

Hosting an llms.txt file within your developer portal’s root directory could assist here. llms.txt is a proposed metadata standard for LLM-based AI crawlers. Although still experimental, it’s designed to function similarly to a robots.txt or sitemap.xml but for LLMs.

Hedge Your Bets on AI Protocols

Although MCP is gaining rapid market share, it’s not the only option on the table. Other AI-to-API protocols have emerged, like Google’s proposed Agent2Agent Protocol (A2A), Agent Communication Protocol (ACP), and others. Unlike MCP, which prioritizes direct agent-to-API communication, alternatives like A2A focus on agent-to-agent coordination or incorporate additional reasoning layers. A developer portal is, by its nature, abstracted and agnostic to underlying shifts in AI protocol and API standards, helping cater to trends as they come and go.

Developer Portals Support Humans And AI

It’s an exciting time for APIs — especially the ones well-positioned for new agentic AI and MCP usage patterns. Developer portals are poised to support this growth, making them an essential element for API publishers in an agentic future. All in all, developer portals are very complementary to both traditional developer and AI agent-focused integration means, offering an indispensable human and machine touchpoint to guide the consumption of APIs at large.