Magec

Magec is a self-hosted platform designed for building and orchestrating multi-agent AI systems that operate entirely on a user's own server. It enables the creation of AI agents, each with its own configurable brain, memory, voice, and tools, allowing users to assemble them as independent units in a matter of minutes. These agents can be chained together into collaborative teams, forming complex workflows that can be visually organized with parallel branches, conditional routing, and loops on a canvas interface.

The platform provides mechanisms for agents to interact with real-world systems through MCPs (which allow operation of external systems) and skills, which are playbooks that guide agents in performing tasks according to user-defined procedures. Magec’s memory feature allows agents to retain long-term information, enabling them to recall facts and context over extended periods and apply this knowledge in future interactions.

Magec supports multiple client interfaces and integrations, making agents accessible through voice commands, Telegram, webhooks, cron jobs, HTTP requests, and even other agents via A2A (agent-to-agent) communication. Each client is assigned a unique token, and access can be restricted to specific agents or workflows. The voice interface allows users to communicate conversationally with agents, who can process spoken commands, perform tasks, and respond using text-to-speech.

Security and privacy are addressed through encrypted secrets management, ensuring that sensitive data such as API keys remain protected and are only exposed to tools at execution time, while models interact with placeholders. Every agent run is fully recorded, with each step, decision, and error timestamped and available for inspection from an admin panel.

Installation is streamlined to a single command, with the platform running fully locally and requiring no external API keys. Magec is aimed at users seeking to build, automate, and manage sophisticated AI agent workflows with complete control over deployment and data, leveraging a self-hosted environment.