agent·interface

The reference site for the agent-interface layer

What is an agent interface?

Definition

An agent interface is the layer where an AI agent meets everything outside the model. It has two sides: the surface a human uses to direct, supervise, and correct the agent — and the protocols the agent uses to operate software, tools, and other agents. Chat panels, approval gates, MCP connections, and computer-use screen control are all parts of it.

updated 2026-07-24 · status: living document · every claim sourced · tracker updated 2026-08-27 · json

Disambiguation: "agent interface" also has an older meaning — the screen a human support agent uses in contact-center software. This site covers the AI sense of the term.

HUMAN IN COMMAND AGENT THE MODEL, ACTING SOFTWARE TOOLS · DATA · AGENTS APPROVALS · MODES · TRAIL MCP · A2A · AG-UI · X402 THE AGENT INTERFACE IS BOTH SEAMS
supervision flows left · capability flows right · lose either and you don't have an agent product

01The short version

9,600+
servers in the official MCP registry
registry.modelcontextprotocol.io
150+
organizations behind the A2A protocol
Linux Foundation, Apr 2026
10 mo
ChatGPT Atlas, launch to shutdown
the graveyard has the story
97%
of llms.txt files never fetched by an AI bot
137k-domain log study

02What does an agent interface do?

An agent interface turns a capable model into a system people can actually delegate to. Downward, it gives the agent structured access to the world: tools it can call, data it can read, software it can operate, other agents it can hand work to. Upward, it gives the human command: what the agent may touch, when it must ask, what it is doing right now, and what it did last Tuesday.

The two directions are one design problem. Every new capability granted below creates a supervision question above — an agent that can spend money needs a spending gate; an agent that can act across apps needs an audit trail that spans them. Products that treat the protocol side and the UX side as separate projects ship the gap between them.

03Agent interface vs. chatbot UI vs. copilot

Chatbot UICopilotAgent interface
System canAnswerSuggest inside one appAct — across tools, over time
Human's jobAsk wellAccept or reject inlineDirect, bound, supervise, audit
Core widgetMessage threadInline completion / diffApproval gate, permission modes, run view, trail
Failure modeWrong answerBad suggestionUnwanted action — which is why the interface is the safety system
Wiring underneathOne model APIApp-internal hooksProtocols: tools, agent-to-agent, frontend, payments

The boundary that matters is action. A system that only answers needs good conversation design. A system that acts needs an agent interface, and the difference is not cosmetic — it is approval gates, permission modes, interruption, and an audit trail, none of which a chat thread provides.

04The protocol layer, mapped

The agent-software side of the interface is where the standardization is happening. These are the protocols and approaches that matter right now — each links to its tracker entry with a current status call and sources:

MCP — Model Context Protocol The standard connector between an agent and its tools and data, JSON-RPC based. Spec revision 2026-07-28 shipped stable, dropping protocol-level sessions for a stateless core. Agentic AI Foundation (Linux Foundation); created at Anthropic · Established A2A — Agent2Agent Agent-to-agent interop: discovery via Agent Cards, task delegation, long-running exchanges. v1.0 shipped March 2026, v1.0.1 in May. Agentic AI Foundation (Linux Foundation) since Aug 2026; TSC: AWS, Cisco, Google, IBM, Microsoft, Salesforce, SAP, ServiceNow; created at Google · Established AG-UI — Agent–User Interaction Protocol Standardizes the agent-to-frontend seam: ~31 typed event kinds over SSE for streaming, state sync, interrupts, and generative UI. CopilotKit — the notable foundation holdout · Rising ACP — Agent Client Protocol (Zed) The "LSP for coding agents": a JSON-RPC protocol connecting any editor to any coding agent. The joint agent registry now lists 42 agents, including Cursor, Devin, GitHub Copilot, Gemini and Codex. Jointly governed by Zed Industries and JetBrains under an interim model that names an independent foundation as the eventual home · Established AP2 — Agent Payments Protocol Cryptographically signed mandates so an agent purchase carries durable, auditable proof of what the human approved. v0.2 adds human-not-present payments. FIDO Alliance (donated by Google, April 2026) · Early x402 HTTP-native machine payments reviving status code 402: a server quotes terms, the agent pays and retries with proof. x402 Foundation (Linux Foundation); created at Coinbase · Rising

The full tracker → is the living version of this map: statuses, stewards, spec links, and what changed, reviewed continuously.

The graveyard

This layer kills fast. We keep the list.

  • OpenAI Operator7 monthsabsorbed into ChatGPT agent
  • ChatGPT Atlas<10 monthsthe standalone agentic browser, discontinued
  • AgentKit Agent Builder8 months to deprecationvisual agent-building lost to code again
  • IBM ACP~5 monthsmerged into A2A; its acronym got recycled

Every death, dated and sourced →

05The human layer: the patterns that keep people in command

No standards body owns the human side yet, but shipped products have converged on a recognizable pattern language. The load-bearing ones:

The ladder, live — same job, four permission modes

06Why agent interfaces matter in 2026

Models crossed the capability line before products crossed the trust line. What limits deployment of agents today is rarely whether the model can do the work — it is whether an organization can predict, bound, and audit the work. That is an interface problem, in both directions: protocols make capability legible to software; supervision UX makes it legible to people.

The teams shipping agents successfully are the ones treating the interface as the product. The model is rented; the interface — the permissions, the gates, the trail, the integrations — is what they actually build.

07Frequently asked questions

Is an agent interface the same thing as a chatbot UI?
No. A chatbot UI presents a conversation; an agent interface governs actions. The moment the system can do things — run code, send email, spend money — the interface must add approval gates, permission modes, progress visibility, and an audit trail. Chat is often the entry point, but it is the smallest part.
What is the difference between MCP and an agent interface?
MCP (Model Context Protocol) is one protocol inside the agent-software layer of the agent interface: it standardizes how an agent connects to tools and data. The agent interface is the whole seam — MCP and its peer protocols below, plus the human-facing control surface above.
What are the main agent-interface protocols?
The most established is MCP for agent-to-tool connections. Agent-to-agent communication and agent-to-frontend streaming have their own emerging protocols, and screen-level control (computer use) is a distinct approach that skips protocols by operating the same interface humans use. The tracker on this site follows each one with a current status call.
What does a good human-agent interface include?
Five recurring elements: an approval gate that shows consequences rather than raw actions, graduated permission modes (read-only, suggest, auto-with-gates, scoped full-auto), live progress the human can interrupt, clean handoff when the agent is stuck, and an append-only audit trail.
Who maintains this site?
Agent Interface (agentinterface.app) is an independent reference site. The tracker and guides are maintained continuously, with every claim linked to a source. The Library is an archive of published verdicts from The AI Council, an earlier multi-model experiment on this domain.

Also on this domain: the Library — an archive of published verdicts from The AI Council, an earlier experiment here in which four frontier models argued real questions and handed down one verdict. It stays up because the deliberations are still worth reading.