xpert-ai/xpert
XpertAI is an open-source platform for building, running, and evolving ai agents, providing extensible capabilities and infrastructure for diverse AI systems.
What it solves
Xpert is an open-source platform designed to build, manage, and govern teams of AI agents. It addresses the complexity of moving from simple chat interfaces to professional AI teams that can execute complex workflows, integrate with enterprise data, and operate within a governed environment with permissions and audit trails.
How it works
Xpert uses a plugin-first, modular architecture based on NestJS. It combines flexible AI reasoning (Agents) with deterministic business processes (Workflows) using LangGraph state graphs.
Key components include:
- Xpert Bosi: A desktop application that acts as a team leader, allowing users to assemble agents, manage files, and execute local shell commands or cloud computer operations.
- Xpert Platform: The backend runtime that manages agent execution, permissions, and shared resources. It supports a plugin system for adding models, tools, and MCP (Model Context Protocol) apps.
- Workbench: Dedicated interactive views where users can inspect, correct, and approve the work produced by AI agents.
- Knowledge Integration: Uses RAG and GraphRAG to parse files and retrieve evidence with citations.
Who it’s for
- Developers building agentic applications and multi-agent workflows.
- Organizations needing a governed, self-hosted AI agent platform for business processes.
- Enterprise users who want to integrate AI agents with their internal data and BI tools.
Highlights
- Hybrid Architecture: Combines agentic reasoning with structured workflows for predictable results.
- Plugin-First Design: Everything from model providers to business apps is a plugin, making the system highly extensible.
- Governed Execution: Includes built-in approvals, typed tools, and audit trails for secure enterprise use.
- Local Execution: The Bosi desktop app allows agents to run local shell commands with per-command approval.
- RAG/GraphRAG Support: Integrated capabilities for parsing files and citing sources for grounded AI responses.
Related
- Project
- Project
- Project
- Project
- Project