Nexent

Nexent is a self-hosted AI agent platform for people and teams who want to build agents by describing tasks in plain language. It generates executable agents without requiring code or a visual workflow editor.

• Agent creation and reuse: Build agents with tools, skills and behavior constraints, choose agents from a marketplace, and restore earlier versions through version history.
• Tools and collaboration: Connect external services through MCP and let agents cooperate on distributed workflows through the Agent-to-Agent (A2A) protocol.
• Persistent memory: Keep context across conversations with separate memory for each user and each user's interactions with an agent.
• Document knowledge and search: Import documents into a knowledge base with automatic summaries and access controls, then combine private material with internet search and citations for checking sources.
• Models and interaction: Connect OpenAI-compatible APIs for language, embeddings, vision, speech recognition and speech synthesis, and interact through text, voice, images or files.

Nexent is open source under the MIT license and runs locally or on private infrastructure using Docker or Kubernetes. Both support x86_64 and ARM64 hardware and require at least four CPU cores. Minimum RAM is 8 GiB for Docker and 16 GiB for Kubernetes. Teams can use tenant isolation and role-based permissions. Self-hosting runs the platform on your infrastructure; external model providers receive model requests when you use them, and internet search requires connectivity.