affaan-m/ECC
ECC is a comprehensive agent harness performance optimization system providing skills, instincts, memory, security, and research-first development for various AI agent harnesses including Claude Co...
astrid-runtime/astrid
Astrid is an innovative operating system focused on providing a secure and portable runtime environment for composable software, particularly AI agents. It addresses the inherent security risks of untrusted agent code by implementing robust, OS-grade boundaries rather than relying on prompt-based trust. The core idea is to treat software components as encapsulated WebAssembly 'capsules,' each with explicit, constrained authority. The `astrid-daemon` kernel is intentionally minimalist, handling event routing, capability enforcement, sandbox execution, and audit trails without incorporating intelligence, model handling, or business logic. This separation ensures that vulnerabilities in capsules cannot compromise the kernel or other components. Astrid employs a multi-faceted security model, including a cryptographic capability system where every resource access (files, network, tools) is governed by signed, scoped, revocable grants; a WASM sandbox (`Wasmtime`) that prevents ambient authority and restricts external effects to capability-checked host calls; and a manifest-based access control system. IPC interactions are also governed by ACLs, and an audit chain provides tamper-evident logging. The system is designed for live capsule lifecycles, allowing installation, upgrades, and removal on a running daemon without restarts. It supports a component model with versioned `astrid:*` WIT packages for well-defined APIs. Astrid explicitly integrates with the Model Context Protocol (MCP) by providing a secure execution environment, managing capability tokens, and offering granular control over resource access, which is crucial for secure and auditable AI agent operations.
https://github.com/astrid-runtime/astrid
Also in Agent Platforms
open-webui/open-webui
Open WebUI is a self-hosted, extensible AI platform supporting multiple LLM runners and OpenAI-compatible APIs with advanced features like Retrieval Augmented Generation, model building, and offline deployment.
Mintplex-Labs/anything-llm
AnythingLLM is an all-in-one AI application with full MCP compatibility that enables intelligent chat interactions with documents using various LLMs, multi-user support, and no-code AI agent building capabilities.
aaif-goose/goose
goose is an open-source, extensible AI agent providing desktop, CLI, and API interfaces for code, workflows, and general tasks, featuring broad LLM provider compatibility and connection to MCP exte...
HKUDS/nanobot
nanobot is an ultra-lightweight, open-source, self-hosted personal AI agent framework in Python, offering a WebUI, tools, memory, MCP integration, multi-agent workflows, automation, and chat app co...
zhayujie/CowAgent
CowAgent is an open-source, multi-model AI assistant and Agent Harness designed for autonomous task planning, tool execution, and continuous learning, featuring native Model Context Protocol (MCP) ...
sickn33/agentic-awesome-skills
This project provides an installable GitHub library and npm installer for over 1,900 agentic skills and playbooks, designed to enhance AI coding assistants like Claude Code, Gemini CLI, Cursor, and...
danny-avila/LibreChat
LibreChat is an open-source enhanced ChatGPT clone that integrates multiple AI models and supports the Model Context Protocol (MCP) for advanced AI assistant features including no-code agents, secure code execution, multimodal interactions, and customizable conversation management.