Claude Code
Anthropic presents Claude Code across terminal, IDE, web, desktop, and other surfaces. LicoUp connects through its CLI stream-json interface.
Official Claude Code pageFive coding agents. Five ways to connect.
Switching coding agents often means switching sessions, repeating context, and losing track of approvals. Their interfaces explain why—and what a collaboration client needs to keep together.
Representative interfaces: Claude Code, Codex, Cursor, GitHub Copilot, and OpenCode.
The interface decides what a collaboration client can see and keep coherent.
An agent interface is simply the way another product exchanges messages, progress, and actions with a client. Some agents stream lines through a terminal process. Others expose structured requests, events, or a local web connection.
Those shapes affect whether a client can continue the right session, show live progress, preserve useful context, and ask for approval around an action. The table keeps the real interface names so technical readers can compare them; the final column explains why each shape matters.
Five representative agent interfaces.
These interfaces connect coding agents to LicoUp in different ways. Check current compatibility and setup requirements before choosing an agent.
| Agent | Product surface | LicoUp connection | What this means for the conversation |
|---|---|---|---|
| Claude Code | CLI | stdio stream-json | Streaming output can carry live progress and lifecycle information from a terminal-native session. |
| Codex | Desktop, CLI, App Server | stdio JSON-RPC App Server | Structured requests and events can make turns, lifecycle changes, and approvals explicit. |
| Cursor | Desktop, CLI | direct CLI process | A process bridge connects an editor-centered product with a separate collaboration surface. |
| GitHub Copilot | CLI, ACP | stdio ACP | A shared agent protocol can carry structured session and capability semantics between client and agent. |
| OpenCode | CLI, Local Server | loopback HTTP + SSE | A local server keeps the connection independent from one terminal process while streaming events back to the client. |
A multi-agent client should normalize without losing semantics.
One agent may emit structured lifecycle evidence, another may stream JSON, another may expose ACP, and another may own a loopback HTTP server. Those differences carry useful information about resume, streaming, replay, cancellation, steering, and runtime identity.
A collaboration client needs an internal contract that can preserve those semantics while still presenting one visible conversation to the user.
LicoUp’s generated inventory explicitly models these interface families so heterogeneous agents can converge on a coherent human-facing experience.
How the five fit into the collaboration layer.
Codex
OpenAI maintains Codex as a coding agent with a terminal CLI and structured local interfaces. LicoUp connects through the stdio App Server JSON-RPC interface.
Official Codex repositoryCursor
Cursor combines editor-native agent workflows with a standalone CLI. LicoUp connects through a direct CLI process.
Official Cursor CLI pageGitHub Copilot CLI
GitHub offers a terminal-native Copilot agent and ACP integration. LicoUp connects through stdio ACP.
Official Copilot CLI pageOpenCode
OpenCode spans terminal, desktop, IDE, and local-server workflows. LicoUp connects through its loopback HTTP + SSE server interface.
Official OpenCode docsFind your connection
LicoUp includes thirteen conversation adapters across CLI, structured-process, local-server, and protocol-native interfaces. Choose an installed runtime that provides the interface its adapter needs.
See all compatibility requirementsInterface details become everyday experience.
The protocol name matters less to most people than the behavior it enables.
Sessions
Resume the intended thread and recognize when an agent starts a fresh one.
Context
Understand which history, files, and instructions a participating agent receives.
Approvals
See the proposed action and decide when local or external effects should proceed.
Start with the agent you already use.
Install the agent runtime and complete its own authentication setup. LicoUp discovers local runtimes and connects through the interface listed above; the agent’s version and capabilities determine the available behavior.
The compatibility overview brings all thirteen adapters and their setup conditions together, including the additional readiness check required by DeepSeek Harness.