plugin architecture

MCP is like a USB-C port for AI applications. Just as USB-C provides a standard way to connect various devices (phones, laptops, cameras) to different peripherals (chargers, monitors, storage), MCP pr

75 Focused Plugins - Granular, single-purpose plugins optimized for minimal token usage and composability" — Shows scalable plugin architecture with 75 isolated, composable plugins following single-re

local plugins...sync automatically across desktop and CLI" — Article exemplifies plugin architecture by showing local plugin installation, automatic synchronization across desktop and CLI interfaces.

Before MCP, every AI coding tool had its own plugin system — ChatGPT Plugins, Cursor extensions, Copilot adapters — none of them compatible with each other. MCP collapses all of that into a single pro

You install a channel as a plugin and configure it with your own credentials. Telegram, Discord, and iMessage are included in the research preview." — Demonstrates plugin-based extensibility model wit

The domino-claude-plugin is designed to be extended. The repo ships 20+ built-in skills, two output styles, and an example hooks directory, all working templates you can fork and tailor to your team's

Think of MCP like a USB-C port for AI applications" — The USB-C analogy explicitly demonstrates MCP as a plugin/interface architecture pattern—a universal connector that allows different components to

Adding settings.json as a distributable component in plugin.json (like commands, agents, skills, hooks, mcpServers) would solve this" — Article proposes extending plugin.json schema to include setting

Adapt pi to your workflow, share your wares with others. No forking, no recompilation. Just push to NPM or git and anyone can benefit" — Article demonstrates extensibility mechanism where users can cr

Your AI coding agent can manage skills for you. All you need is this managing-skills skill!" — Demonstrates concrete implementation where agents use a skill management system to dynamically add/remove

makes it actually easy to publish, discover and install extensions" — Article demonstrates improved developer experience for plugin lifecycle (publishing, discovery, installation) through thoughtful p

import skills from @openclaw hub & @vercel" — LettaBot's skill marketplace allows importing capabilities from external hubs, enabling extensibility

Instead of hardcoding integrations, you build an MCP Server that says, "Hey, here are the tools I have, and here is how you use them." The AI model then reads that "menu" and decides when to order off

Claude Code can browse and install servers from the official MCP registry. The registry is fetched at startup and updated automatically." — The MCP registry demonstrates a practical implementation of

Anthropic is building a platform, not just a chat wrapper. The plugin architecture is what makes this interesting beyond the core feature set." — Claude Code implements an open, inspectable plugin sys

A new claude code plugin we built that enables two use cases" — Demonstrates plugin-based extension approach to enhance Claude Code capabilities with new functionality

If an MCP connector for a service doesn't exist, in some cases, it's possible to get Claude to create one for you and run your own MCP server to connect to that service." — Demonstrates MCP as a plugi

ChatGPT apps directly in the terminal" — The ability to load external ChatGPT apps into Codex terminal demonstrates a plugin-based architecture where third-party applications can extend core functiona

Tiny core, able to write plugins for itself as you use it." — Pi exemplifies dynamic plugin generation where an agent autonomously creates functionality to extend its own capabilities, enabling self-d

New /claude-api skill, InstructionsLoaded hook, ${CLAUDE_SKILL_DIR} variable" — Article documents new developer features that demonstrate skill architecture patterns: dedicated /claude-api skill, hook

Bundle your extensions, skills, prompt templates, and skills as pi packages, distribute via NPM or git" — Shows practical plugin distribution mechanism through standard package managers (NPM/git), ena

I asked it to create a hook." — Demonstrates practical use of hook mechanism as a real implementation pattern for extending agent behavior.

You can already do status line, widget above editor, full editor replacement, custom message rendering, fully custom UI for commands. Now you can swap out the full footer and do whatever you want." —

Bundle your extensions, skills, prompt templates, and skills as pi packages, distribute via NPM or git" — Article demonstrates practical implementation of extensible plugin architecture through pi's p

We're bringing full feature parity with MacOS: file access, multi-step task execution, plugins, and MCP connectors." — Cowork implements plugins as a standard feature across platforms, showing practic

full feature parity with MacOS: file access, multi-step task execution, plugins, and MCP connectors" — Plugins are a primary feature being launched on Windows platform

claude just hangs if you forget to include the owner attributes in the marketplace.json of a plugin" — Direct evidence of a specific plugin configuration requirement (owner attributes in marketplace.j

wanted to bring the compound engineering plugin that @kieranklaassen made for claude code over to pi" — Concrete example of cross-platform plugin portability and reuse patterns in AI-assisted developm

So hooks won't inherit OTEL_* env vars from settings.json anymore?" — Article questions removal of environment variable inheritance for hooks, which breaks existing observability patterns

even those that not officially supported" — Article specifically addresses integrating unofficial/unsupported tools, demonstrating extensibility beyond base integrations

Modularity: Easily extendable to support multiple MCP Servers." — MCP's architecture explicitly supports modularity and extensibility, allowing multiple independent servers to be connected without mod

[DIRECT] "Model Context Protocol servers" — MCP servers function as a plugin architecture — they are independently developed components that extend LLM capabilities through a standardized protocol.

install with pi install npm:pi-monitor" — Demonstrates plugin architecture pattern where agent capabilities are extended through npm-installable modules, showing modularity and extensibility in agent

[DIRECT] "Hooks are a killer feature" — Hooks feature enables extensibility in pi agent, core plugin/extension pattern

Upload your skills and use vector search to find new ones" — Article showcases skill sharing and discovery mechanism as core feature

it uses your existing Claude Code plugins and setup" — TinyClaw leverages an existing plugin architecture, demonstrating that lightweight agent implementations can maintain compatibility with establis

Easy to build your own voice agent using Agent Skills and Pi Extensions." — Pi Extensions provides a concrete plugin/extension mechanism for customizing agent behavior

Hot-reload everything via /reload" — System supports dynamic hot-reload of extensions without restart, enabling rapid development iteration

had the clanker build an extension that implements a minimal tool output mode" — Article demonstrates practical use of extension system to add feature-specific functionality without modifying core

Built extension for tool approvals to make pi more appealing for non-techies and DevOps people" — Article demonstrates practical extension mechanism built for the pi coding agent, showing extensibilit

[DIRECT] "claude just hangs if you forget to include the owner attributes in the marketplace.json of a plugin" — Demonstrates a specific plugin configuration requirement (owner attributes in marketpla

[INFERRED] "This might end up being one of the largest contribution PRs to the repo so far 👀" — The scale of contribution (43 skills, 6 subagents) demonstrates a modular framework design that support

Plugin APIs are a form of extensibility architecture, but this tweet reveals nothing about how context flows through plugins or how they manage state.