Agentic AI has been everywhere in 2025. Forrester called it one of the top emerging technologies of 2025. Forbes has called it “the next big breakthrough that’s transforming business and technology.” BizTech Magazine discussed agentic AI “revolutionizing business and daily life.”
Agentic AI only seems to be spreading. According to Cloudera’s recent The Future of Enterprise AI Agents report, 57% of the companies they surveyed had already implemented agentic AI. As if that weren’t already proof enough, 96% of respondents plan on expanding their use of agentic AI in the next 12 months.
The rapid rise in agentic AI’s popularity has caused numerous developers and designers to create a number of new standards that allow AI agents to interact directly with APIs. They’re not all competing, either — some are complementary.
To help you learn more, we’ve compared five emerging AI agent-to-API standards to help you understand the landscape and decide which one’s right for for you.
Anthropic MCPMCP, which stands for Model Context Protocol, is Anthropic’s solution for standardizing how data and tools are supplied to an AI tool like an LLM. Its main goal is to extend LLMs by connecting applications, documents, APIs, and other data sources to add context through a common interface. Its primary purpose is to provide AI tools like Claude for Desktop or Cursor with relevant data for a particular task.
Example MCP server:
{ "mcpServers": { "brave-search": { "command": "docker", "args": [ "run", "-i", "--rm", "-e", "BRAVE_API_KEY", "mcp/brave-search" ], "env": { "BRAVE_API_KEY": "YOUR_API_KEY_HERE" } } } }Who’s using MCP:
Google’s A2A, which stands for Agent2Agent, is a protocol for allowing two or more opaque AI agents to communicate with one another. Instead of a one-size-fits-all framework for every situation, A2A allows agents to assign tasks across boundaries, whatever they might be. Its main purpose is to enable agents to communicate and interact with one another, using “Agent Cards” to describe each agent’s capabilities.
Example of A2A:
{ "name": "GeoSpatial Route Planner Agent", "description": "Provides advanced route planning, traffic analysis, and custom map generation services. This agent can calculate optimal routes, estimate travel times considering real-time traffic, and create personalized maps with points of interest.", "url": "https://georoute-agent.example.com/a2a/v1", "provider": { "organization": "Example Geo Services Inc.", "url": "https://www.examplegeoservices.com" }, "version": "1.2.0", "documentationUrl": "https://docs.examplegeoservices.com/georoute-agent/api", "capabilities": { "streaming": true, "pushNotifications": true, "stateTransitionHistory": false }, "authentication": { "schemes": ["OAuth2"], "credentials": "{\"authorizationUrl\": \"https://auth.examplegeoservices.com/authorize\", \"tokenUrl\": \"https://auth.examplegeoservices.com/token\", \"scopes\": {\"route:plan\": \"Allows planning new routes.\", \"map:custom\": \"Allows creating and managing custom maps.\"}}" }, "defaultInputModes": ["application/json", "text/plain"], "defaultOutputModes": ["application/json", "image/png"], "skills": [ { "id": "route-optimizer-traffic", "name": "Traffic-Aware Route Optimizer", "description": "Calculates the optimal driving route between two or more locations, taking into account real-time traffic conditions, road closures, and user preferences (e.g., avoid tolls, prefer highways).", "tags": ["maps", "routing", "navigation", "directions", "traffic"], "examples": [ "Plan a route from '1600 Amphitheatre Parkway, Mountain View, CA' to 'San Francisco International Airport' avoiding tolls.", "{\"origin\": {\"lat\": 37.422, \"lng\": -122.084}, \"destination\": {\"lat\": 37.7749, \"lng\": -122.4194}, \"preferences\": [\"avoid_ferries\"]}" ], "inputModes": ["application/json", "text/plain"], "outputModes": [ "application/json", "application/vnd.geo+json", "text/html" ] }, { "id": "custom-map-generator", "name": "Personalized Map Generator", "description": "Creates custom map images or interactive map views based on user-defined points of interest, routes, and style preferences. Can overlay data layers.", "tags": ["maps", "customization", "visualization", "cartography"], "examples": [ "Generate a map of my upcoming road trip with all planned stops highlighted.", "Show me a map visualizing all coffee shops within a 1-mile radius of my current location." ], "inputModes": ["application/json"], "outputModes": [ "image/png", "image/jpeg", "application/json", "text/html" ] } ] }Who’s using A2A:
Agent Connect Protocol (ACP) and Agent Gateway Protocol (AGP) are both specifications created under the AGNTCY initiative, an open standardization created by CISCO, LangChain, and Galileo, among others. They’re more than just another set of API specifications. Instead, they’re intended to enable an “Internet of Agents,” an open-ended ecosystem allowing autonomous agents, AI services, and traditional software to interact. Agent Connect Protocol (ACP) is designed to describe a singular agent, making it more discoverable and easier to automate.
Example ACP specification:
{ "metadata": { "name": "org.agntcy.mailcomposer", "version": "0.0.1", "description": "Compose email for campaigns" }, "capabilities": { "threads": false, "interrupts": false, "callbacks": false }, "input": { "$defs": { "Message": { "properties": { "type": { "$ref": "#/$defs/Type", "description": "Indicates the originator of the message, a human or an assistant" }, "content": { "description": "The content of the message", "title": "Content", "type": "string" } }, "required": ["type", "content"], "title": "Message", "type": "object" }, "Type": { "enum": ["human", "assistant", "ai"], "title": "Type", "type": "string" } }, "properties": { "messages": { "anyOf": [ { "items": { "$ref": "#/$defs/Message" }, "type": "array" }, { "type": "null" } ], "default": null, "title": "Messages" } }, "required": ["messages"], "type": "object" } }Agent Gateway Protocol (AGP) extends ACP by adding a higher-level routing and acceleration layer. It’s essentially an ACP gateway, exposing ACP-compatible agents through a single endpoint. AGP allows all available agents to be listed and filtered in a variety of ways.
Example AGP specification:
{ "type": "agp/task/start", "task": { "id": "task-12345", "title": "Generate Weekly Report", "description": "Compile and summarize weekly sales data.", "context": [ { "type": "text/plain", "content": "Sales data for the week of May 5-11 is attached." }, { "type": "application/json", "content": { "sales": [ {"region": "North", "total": 15000}, {"region": "South", "total": 12000} ] } } ] } }Who’s Using AGNTCY’s ACP/AGP:
Wildcard’s agents.json standard is a lightweight schema built on top of the OpenAPI standard for describing agents in a format that’s easily consumed and standardized. Every agents.json includes metadata about the agent, including its name, function, authentication requirements, and expected inputs and outputs. This allows tools, front-ends, and SDKs to discover and implement the right agent. It also allows developers to define tools and flows, making it fast and easy to switch out resources. It’s an ideal bridge between AI-driven tools like LLMs and APIs without creating new tools for each service.
Example agents.json:
[ { "id": "assistant", "name": "General Assistant", "description": "Helps with a wide range of questions and tasks, including summarization, writing, and general reasoning.", "system_prompt": "You are a helpful assistant capable of performing general-purpose tasks.", "tools": ["search_web", "summarize_text", "generate_email"] }, { "id": "calendar", "name": "Calendar Bot", "description": "Manages events and appointments. Can schedule, update, and delete calendar events.", "system_prompt": "You are an expert in managing personal schedules and calendars.", "tools": ["create_event", "update_event", "delete_event", "list_events"] } ]Who’s using agents.json:
LangChain’s Agent Protocol is a framework-agnostic, REST-based specification for deploying, orchestrating, and monitoring LLM-powered agents. It translates common functions agents perform, like planning, memory, and step-based executions, into a consistent API format. This allows agent frameworks like LangGraph, AutoGen, or CrewAI to extend their memory, workflows, and execution. It’s intended to be used in production environments or as a backend for UIs, orchestration engines, or logging systems. By acting as an abstraction layer, it helps orchestrate open-source agents into a cohesive ecosystem.
Example Agent-Protocol:
"openapi": "3.1.0", "info": { "title": "Agent Protocol", "version": "0.1.6" }, "tags": [ { "name": "Runs", "description": "A run is an invocation of an agent, optionally, on a thread. If applied to a thread, it updates the state of the thread. Otherwise, it has not state or memory persistence." },Who’s using Agent-Protocol:
Agile also created its own agent specification for standardizing and describing agents. Unlike the other AI Agent-to-API-standards on our list, Agile’s Agent Specification is available as a YAML file, making it ideal for local installations, production environments, and integrating into programming languages like Python. It’s also more detailed than the other standards we’ve mentioned, with fields describing each agent’s domain, value, level of agency, and even the ideal target user. These detailed descriptions make it even easier for an agentic ecosystem to quickly find the right agent for a particular task.
Example Agent Specification:
id: "urn:agent:healthcare:diagnostic_assistant:3" name: "Diagnostic Assistant" fullyQualifiedName: "healthcare.ai.diagnostic_assistant" description: "Autonomous agent supporting medical diagnostics and patient care recommendations." domain: "healthcare" version: "3.5.0" environment: "staging" agentOwner: "[email protected]" status: "ACTIVE" kind: "Single Agent" agentGoal: "Provide accurate diagnostic recommendations to clinicians while tracking patient outcomes." targetUser: customer valueGeneration: "DecisionMaking" interactionMode: MultiTurnConversation runMode: "Reactive" agencyLevel: "ModelDrivenWorkflow" toolsUse: true learningCapability: "Fine Tuning" Final Thoughts on AI Agent-to-API StandardsMuch like AI itself, agentic AI isn’t going anywhere. If anything, it’s only going to become more prevalent, as AI needs to be able to execute changes and perform functions autonomously. AI agent-to-API standards are an essential part of that process, as they allow agents to communicate with one another safely, securely, and efficiently.
It remains to be seen if all of these standards will stick, as we saw with the rise of the OpenAPI specification, or if some will fall by the wayside. However, many already seem here to stay since they can be used together to create a robust, efficient ecosystem. MCP is already making its mark on the tech landscape, for instance, judging from its already widespread adoption. MCP lacks a discoverability layer, though, as Anthropic designed the standard assuming you’d already be familiar with each server you’re using. This means MCP may work best in conjunction with standards like Wildcard’s agents.json or LangChain’s Agent-Protocol.
If you’re hoping to integrate AI agents into local tools like Claude for Desktop or Cursor, Anthropic’s MCP standard will suit your needs. If you’re designing a multi-agent ecosystem, try Google’s A2A or AGNTCY’s ACP and AGP standards. If you want a lightweight solution to easily connect agents to existing APIs, try Wildcard’s agents.json. If you hope to integrate agents into your development environment, try LangChain’s Agent Protocol or Agile Lab’s Agent Specification.
All Rights Reserved. Copyright , Central Coast Communications, Inc.