Remote Frog

The naming chaos: agents, skills, plugins… what’s what?

No charm for the people who have issues remembering names. Sigh. Every AI platform use different names. But if you scratch under the surface, the pattern is visible:

• Main entity: the “assistant” or “bot” you configure for a purpose.
• Extra capabilities: the tools, plugins, or connectors that extend what the assistant can do.

Here’s how each platform calls them:

Platform	Main entity	Extra capabilities
OpenAI (ChatGPT)	GPTs (custom GPTs)	Plugins
LibreChat	Agents	Presets (saved configs)
Claude	Projects / Role Agents	Skills
Google Gemini	—	Extensions
Microsoft Copilot	Agents (via Copilot Studio)	Connectors
Other agentic platforms	Agents	Tools / Actions / Connectors

Quick breakdown:

• GPTs are mini-assistants you customize with instructions, files, and tools. Plugins are external services ChatGPT can call (Expedia, Zapier, etc.).
• LibreChat agents are specialized bots with their own tools. Presets are just saved configurations to start chats faster.
• Claude skills are modular, reusable capabilities—like functions the assistant calls when needed.
• Gemini extensions connect to Google services (YouTube, Maps, Gmail…).
• Copilot agents are custom assistants; connectors link them to external data (SharePoint, Salesforce…).

OpenAI: GPTs and Plugins

GPTs (custom GPTs) are small, customized assistants that live inside ChatGPT. You configure them with instructions, files, and tools for a specific purpose—like a “balonmano expert” GPT or a “company helpdesk” GPT. They’re self-contained and shareable.

Plugins are an older concept: external apps and services that ChatGPT can call via API. Think Expedia, Zapier, or any tool that gives ChatGPT access to real-time data or actions like booking flights, searching databases, or integrating with your stack.

LibreChat: Presets and Agents

Presets are saved chat configurations—model, temperature, system prompt, and so on. They’re basically “saved profiles” so you can jump into conversations with the same settings without reconfiguring every time, you can also add a bit of context and a default prompt.

Agents are the more advanced option: specialized assistants with their own tools, configuration, and behavior. Conceptually similar to OpenAI’s custom GPTs—a bot built for a purpose, with extra capabilities baked in.

Claude: Skills

Skills are modular capabilities with defined inputs, outputs, and permissions. Claude calls them when needed—for example, to summarize documents, hit an internal API, or generate a report. Think of them as reusable building blocks for repetitive tasks. They behave like “functions” or plugin-style tools that a Claude assistant can invoke on demand.

Google Gemini: Extensions

Extensions let Gemini call other Google services directly from the chat—YouTube, Maps, Gmail, Drive, Flights, you name it. They work like plugins.

Microsoft Copilot: Agents

Agents are custom assistants you build in Copilot Studio. You give them their own instructions, data sources, and actions—conceptually similar to LibreChat agents or Claude Projects.

Other agentic platforms

Most use Agents (or AI agents) as the main “assistant” entity. Then they add tools, actions, connectors, extensions, or skills as the extra capabilities that those agents can call.

Different names, same architecture: an assistant that knows when to reach for help.