Custom Middleware And Extension System For Agent Behavior Customization

via “agent definition and configuration with role-based context”

Stateful AI agent platform — long-term memory, workflow execution, persistent sessions.

Unique: Treats agent definitions as first-class configuration objects that persist independently of sessions, enabling reusable agent personas with consistent behavior across multiple concurrent conversations

vs others: Cleaner separation of agent configuration from session state compared to frameworks like LangChain where agent setup is often mixed with conversation logic

via “middleware pipeline with pre/post-processing hooks for agent execution”

An open-source long-horizon SuperAgent harness that researches, codes, and creates. With the help of sandboxes, memories, tools, skill, subagents and message gateway, it handles different levels of tasks that could take minutes to hours.

Unique: Implements a composable middleware pipeline with pre/post-processing hooks at multiple execution stages, enabling clean separation of concerns. Middleware can modify execution context, inject additional data, or short-circuit execution, providing fine-grained control over agent behavior.

vs others: More flexible than monolithic agent code because concerns are separated into reusable middleware. More practical than aspect-oriented programming because middleware is explicit and easy to understand.

via “plugin system with administrative and behavioral plugins”

CowAgent (chatgpt-on-wechat) 是基于大模型的超级AI助理，能主动思考和任务规划、访问操作系统和外部资源、创造和执行Skills、通过长期记忆和知识库不断成长，比OpenClaw更轻量和便捷。同时支持微信、飞书、钉钉、企微、QQ、公众号、网页等接入，可选择DeepSeek/OpenAI/Claude/Gemini/ MiniMax/Qwen/GLM/LinkAI，能处理文本、语音、图片和文件，可快速搭建个人AI助理和企业数字员工。

Unique: Implements a hook-based plugin system with defined extension points (pre-processing, post-processing, tool invocation) that allows plugins to intercept and modify the message pipeline without subclassing

vs others: More flexible than configuration-based customization because plugins can execute arbitrary code; more lightweight than full framework extensions because plugins are loaded dynamically at startup

via “agent lifecycle hooks and custom extension points”

Multi-agent platform with distributed deployment.

Unique: Provides a comprehensive hook system covering agent lifecycle points (reasoning, tool execution, error, completion) with access to agent state and ability to modify behavior, enabling custom extensions without modifying core agent code or using middleware.

vs others: More granular than middleware-only approaches because hooks cover agent-level lifecycle; more flexible than fixed extension points because hooks are declaratively registered and can be added/removed at runtime.

via “extensibility framework for custom operations and protocol features”

Agent2Agent (A2A) is an open protocol enabling communication and interoperability between opaque agentic applications.

Unique: Defines a formal extension mechanism at the protocol level (declared in AgentCard, negotiated at discovery) rather than relying on ad-hoc custom fields, enabling controlled extensibility that doesn't fragment the ecosystem

vs others: More structured than uncontrolled custom fields and more discoverable than hidden implementation-specific features, providing a standardized way to extend A2A without breaking compatibility

via “extension system with custom hooks and configuration variables”

An open-source AI agent that brings the power of Gemini directly into your terminal.

Unique: Implements a hook-based extension system where custom JavaScript/TypeScript modules can intercept and modify agent behavior at multiple lifecycle points (pre-prompt, post-response, tool-execution). Variables are interpolated from configuration and environment.

vs others: More flexible than hardcoded customization because extensions can be developed independently and composed together, enabling teams to build complex customizations without modifying core code.

via “middleware-based tool execution pipeline with custom interceptors”

Agent harness built with LangChain and LangGraph. Equipped with a planning tool, a filesystem backend, and the ability to spawn subagents - well-equipped to handle complex agentic tasks.

Unique: Middleware system operates at the LangGraph node level rather than as a wrapper around tool calls, enabling state-aware interception and result eviction without re-executing the agent's reasoning loop. Supports custom handlers that can modify, reject, or transform tool results before they're fed back to the LLM.

vs others: More flexible than tool-wrapping approaches because middleware can access full agent state and modify execution flow, whereas simple tool decorators only see individual tool invocations in isolation.

via “agent-skill-customization-and-specialized-agent-personas”

AI chat features powered by Copilot

via “middleware pipeline for observability and custom logic injection”

The fullstack MCP framework to develop MCP Apps for ChatGPT / Claude & MCP Servers for AI Agents.

Unique: Provides composable middleware pipeline with execution context passing, enabling clean separation of concerns between core agent logic and observability/validation concerns. Middleware can modify execution flow (e.g., skip tool invocation, retry with different parameters) without agent code changes.

vs others: More flexible than decorator-based logging; middleware can access full execution context and modify behavior, enabling sophisticated observability and custom logic injection patterns.

via “middleware pipeline for tool invocation interception and transformation”

The fullstack MCP framework to develop MCP Apps for ChatGPT / Claude & MCP Servers for AI Agents.

Unique: Middleware pipeline operates at the tool invocation level rather than the HTTP/transport level, allowing inspection and transformation of semantic tool calls rather than raw protocol messages; middleware is composable and can be added/removed at runtime without restarting agents.

vs others: More powerful than logging decorators because middleware can modify requests/responses, not just observe them; more maintainable than scattered instrumentation because cross-cutting concerns are centralized in middleware.

via “modular-component-system-capability-extension”

[GenAI Application Development Framework] 🚀 Build GenAI application quick and easy 💬 Easy to interact with GenAI agent in code using structure data and chained-calls syntax 🧩 Use Event-Driven Flow *TriggerFlow* to manage complex GenAI working logic 🔀 Switch to any model without rewrite applicat

Unique: Implements a ComponentSystem where agent functionality is extended through pluggable components (EventListener, Tool, Role) registered with agents rather than subclassing, with components coordinating through a shared RuntimeContext, enabling true composition-based agent design.

vs others: More flexible than LangChain's tool binding (which is function-focused) and cleaner than LlamaIndex's agent subclassing approach, with explicit component types (EventListener, Tool, Role) making intent clearer and enabling better code organization.

via “middleware-based request/response processing pipeline”

A framework for developing applications powered by language models.

via “extensible agent architecture with custom agent creation”

TradingAgents: Multi-Agents LLM Financial Trading Framework

Unique: Provides extensible agent architecture where custom agents can be created by extending base classes and implementing agent-specific logic, then registered in LangGraph graph. Agents receive state as input and produce outputs added to shared state, enabling seamless integration without modifying core framework.

vs others: More extensible than fixed-agent systems because it allows adding custom agents without framework changes. More flexible than generic agent frameworks because it provides trading-specific base classes and patterns that reduce boilerplate for financial agents.

Bindu: Turn any AI agent into a living microservice - interoperable, observable, composable.

Unique: Provides a pluggable extension system with hooks into agent initialization, task execution, and communication, enabling developers to add custom logic without modifying framework code.

vs others: More extensible than monolithic agent frameworks because extensions can be composed and combined to add new capabilities without forking the codebase.

via “custom agent creation with flexible system prompts and tool binding”

Multi-agent framework with diversity of agents

Unique: Provides a flexible agent abstraction where behavior is defined through composition of system prompts, tool registries, and reply generators rather than rigid class hierarchies. Agents can be created declaratively through configuration or programmatically through subclassing, enabling both low-code and advanced customization.

vs others: More flexible than LangChain's agent abstractions because agents are defined through prompts and tool bindings rather than requiring subclassing, and more powerful than simple prompt templates because agents maintain state, manage conversation history, and coordinate with other agents

via “custom agent behavior through inheritance and overrides”

Framework for orchestrating role-playing agents

Unique: Enables low-level customization through class inheritance and method overrides, allowing developers to modify core agent behavior while maintaining crew integration

vs others: More flexible than configuration-based customization but requires more expertise than role-based agent definition

via “extension system for custom agent behaviors and integrations”

Platform for AI-powered software engineers

Unique: Provides a plugin architecture for extending agent behaviors and integrations without core code modification. Extensions hook into the agent execution pipeline, tool registry, and event system, enabling deep customization.

vs others: Offers more extensibility than monolithic agents, while the plugin architecture provides better isolation than monkey-patching.

via “open-source agentic framework with community extensibility”

AI coding dream team of agents for VS Code. Claude Code + openai Codex collaborate in brainstorm mode, debate solutions, and synthesize the best approach for your code.

Unique: Provides fully open-source TypeScript implementation of multi-agent agentic coding framework, enabling community-driven extensions for custom agent roles, debate strategies, and LLM provider integrations. Modular architecture allows deep customization without forking.

vs others: More extensible than proprietary solutions (Copilot, Codeium) because the full source code is available for customization, and more community-driven than closed-source alternatives, though with less guaranteed support and stability.

via “extensible plugin architecture for custom agents”

[COLM 2024] OpenAgents: An Open Platform for Language Agents in the Wild

Unique: Uses a 'one agent, one folder' directory structure with automatic plugin discovery and shared adapters, enabling developers to add custom agents by implementing a standard interface without modifying core code

vs others: More modular than monolithic frameworks but requires more boilerplate than decorator-based plugins; enables code reuse through shared adapters but less flexible than fully composable agent patterns

via “modular agent behavior customization”

Show HN: AgentSwarms – free hands-on playground to learn agentic AI, no setup required!

Unique: The modular approach allows for unprecedented flexibility in defining agent behaviors, unlike rigid frameworks that limit customization.

vs others: Offers greater flexibility than many traditional AI frameworks, which often require extensive coding for behavior changes.