# Nexent

> Open-source zero-code platform for generating AI agents from natural language, with built-in tools, skills, memory, and orchestration.

Nexent is an open-source platform from ModelEngine-Group for building AI agents without code. Users describe what they want in natural language, and the platform generates an executable agent with tools, skills, memory, and orchestration. It can be self-hosted with Docker or Kubernetes and is released under the MIT License. The latest release listed is v2.7.0.

## What It Is

Nexent is a zero-code agent platform built on what its README calls Harness Engineering principles: unified tools, skills, memory, and orchestration with built-in constraints, feedback loops, and control planes. Rather than drag-and-drop workflow design, the user writes a plain-language requirement and the platform produces an agent. Documentation is available in English and Chinese.

## Agent Capabilities

- Multi-model integration: OpenAI-compatible with any provider, covering LLM, embedding, VLM, speech-to-text, and text-to-speech models.
- A2A (Agent-to-Agent) protocol for multi-agent cooperation and distributed workflows.
- Two-tier memory, at user level and user-agent level, for persistent context across conversations.
- Progressive skill disclosure, which loads skills dynamically into context to use the context window efficiently.
- A knowledge base with import and retrieval for 20+ document formats, automatic summaries, and fine-grained access control.
- MCP tool ecosystem for custom and third-party MCP services.
- Internet search blended with private data, with citations and source verification.
- Multimodal interaction through voice, text, images, and files.
- Agent version management with history rollback, and an agent marketplace with official and community agents.
- Multi-tenancy with role-based access control.

## Deployment Model

Docker deployment is recommended for individuals and small teams and runs through `deploy.sh docker`. Kubernetes deployment, installed through Helm, targets enterprise setups needing high availability. The README lists minimum requirements for Docker of 4 CPU cores, 8 GiB memory, and 40 GiB disk, and for Kubernetes of 4 cores, 16 GiB memory, and 100 GiB disk. Offline image packages can be built for air-gapped installs.

## Update: Nexent v2.0

The README announces Nexent v2.0 as an upgrade from v1.0. It added A2A protocol support, progressive skill disclosure, layered memory, multi-tenant user management, agent version management, and the agent marketplace.

## Features
- Zero-code agent generation from natural language
- Multi-model integration (LLM, embedding, VLM, STT, TTS)
- A2A agent collaboration
- Layered user-level and user-agent-level memory
- Progressive skill disclosure
- Knowledge base for 20+ document formats
- MCP tool ecosystem
- Internet search integration
- Citations and source traceability
- Multimodal interaction
- Agent version management
- Agent marketplace
- Multi-tenancy and RBAC
- Docker and Kubernetes deployment
- Offline image packages

## Integrations
MCP, OpenAI-compatible providers, Docker, Kubernetes, Helm, Elasticsearch, PostgreSQL, Redis, MinIO, Supabase

## Platforms
LINUX, WEB, API

## Pricing
Open Source

## Version
v2.7.0

## Links
- Website: http://modelengine-group.github.io/nexent/
- Documentation: https://modelengine-group.github.io/nexent/en/getting-started/overview.html
- Repository: https://github.com/ModelEngine-Group/nexent
- EveryDev.ai: https://www.everydev.ai/tools/nexent