Protocol Taxonomy
28 protocols across 6 categories covering the full agent lifecycle.
โ๏ธ Harness (3)
Frameworks and runtimes for running agents โ execution environments, lifecycle management, and agent scaffolding.
Agent Protocol
Universal open standard for interacting with AI agents, providing a common interface for agent communication regardless of underlying framework.
View Spec →LangGraph
Framework for building stateful multi-actor agent applications as computational graphs with cycles and branching.
View Spec →OpenAI Assistants API
Thread-based agent harness with tool execution, file handling, and persistent conversation management.
View Spec →๐ก๏ธ Safety (3)
Guardrails, alignment techniques, and security frameworks that keep agents operating within safe boundaries.
Constitutional AI
Self-supervised alignment method where AI systems evaluate and revise outputs against a set of principles (a constitution).
View Spec →Guardrails AI
Framework for adding structural, type, and quality guarantees to LLM outputs with validators and corrective actions.
View Spec →OWASP LLM Top 10
Security framework identifying top vulnerabilities in LLM applications including prompt injection, data leakage, and insecure output handling.
View Spec →๐ง Tool Calling (7)
Protocols for agents to invoke external tools, APIs, and data sources with structured inputs and outputs.
CopilotKit/CopilotKit
The Frontend Stack for Agents & Generative UI. React + Angular. Makers of the AG-UI Protocol
View Spec →Model Context Protocol
Open protocol for connecting AI models to external data sources and tools, providing a standardized way to share context.
View Spec →Mouseww/anything-analyzer
ๅ จ่ฝๅ่ฎฎๅๆๅทฅๅ ท๏ผๆต่งๅจๆๅ + MITM ไปฃ็ + ๆ็บนไผช่ฃ + AI ๅๆ + MCP Server ๆ ็ผๅฏนๆฅ AI Agent/IDE | All-in-one protocol analysis toolkit โ built-in browser capture, MITM proxy, JS hooks, fingerprint spoofing, AI analysis & MCP server for agent integration
View Spec →OpenAI Function Calling
Schema-based tool invocation where models output structured JSON matching declared function signatures.
View Spec →Tool Use (Anthropic)
Native tool use capability allowing Claude to call external tools with structured inputs and process results.
View Spec →getsentry/XcodeBuildMCP
A Model Context Protocol (MCP) server and CLI that provides tools for agent use when working on iOS and macOS projects.
View Spec →openclaw/acpx
Headless CLI client for stateful Agent Client Protocol (ACP) sessions
View Spec →๐ง Memory (3)
Systems for agent memory persistence, context management, summarization, and long-term knowledge retention.
LangChain Memory
Pluggable memory subsystem supporting buffer, summary, entity, and knowledge-graph memory backends for conversation persistence.
View Spec →MemGPT / Letta
Virtual context management system that gives LLMs the ability to manage their own memory hierarchy, enabling unlimited context through paging.
View Spec →Zep
Long-term memory service for AI assistants with automatic summarization, entity extraction, and temporal awareness.
View Spec →๐ฏ Orchestration (6)
Multi-agent coordination patterns โ task delegation, group conversations, swarm behaviors, and workflow management.
AutoGen
Multi-agent conversation framework enabling groups of LLM agents to solve tasks through automated chat with customizable interaction patterns.
View Spec →CrewAI
Role-based multi-agent orchestration framework where specialized agents collaborate on complex tasks with defined roles and goals.
View Spec →Swarm (OpenAI)
Lightweight multi-agent orchestration with handoffs and routines, emphasizing simplicity and agent-to-agent transfers.
View Spec →mobile-next/mobile-mcp
Model Context Protocol Server for Mobile Automation and Scraping (iOS, Android, Emulators, Simulators and Real Devices)
View Spec →open-gitagent/gitagent-protocol
A framework-agnostic, git-native standard for defining AI agents
View Spec →๐ก Communication (6)
Agent-to-agent messaging standards โ discovery, negotiation, speech acts, and interoperability protocols.
Agent Communication Language (FIPA ACL)
IEEE FIPA standard for inter-agent communication with performatives like inform, request, propose, and structured message envelopes.
View Spec →Agent Network Protocol (ANP)
Decentralized protocol for autonomous agent networking using DIDs and verifiable credentials for trustless agent-to-agent communication.
View Spec →Agent-to-Agent Protocol (A2A)
Google-led open protocol for agent-to-agent communication enabling discovery, negotiation, and task delegation between heterogeneous agent systems.
View Spec →PennyroyalTea/gibberlink
Two conversational AI agents switching from English to sound-level protocol after confirming they are both AI agents
View Spec →RAIT-09/obsidian-agent-client
Bring AI agents into Obsidian via Agent Client Protocol (ACP), such as Claude Code, Codex and Gemini CLI.
View Spec →machinepulse-ai/world2agent
World2Agent(W2A) is an open protocol that standardizes how Al agents perceive the real world.
View Spec →