English | 简体中文
Xpert is an open-source Agent platform where models, integrations, tools, middleware, Skills,
Assistants, Workbench views, MCP Apps, and complete Agentic Apps share one governed plugin system.
Explore the Plugin Marketplace · Browse Plugin Source · Develop with Agent Skills · Self-host Xpert · Documentation
The Xpert core provides the runtime, contracts, security boundaries, and lifecycle controls. Plugins provide the capabilities and applications that users install: from a model provider or tool to a complete Assistant, multi-Agent workflow, or business-facing Agentic App.
| Discover | Learn | Build |
|---|---|---|
| Find and install capabilities and applications in the Plugin Marketplace. | Study official and community implementations in xpert-plugins. |
Give a coding agent the xpert-plugin-development skill to create, test, deploy, version, and prepare plugins for release. |
This is a live Agent Studio configuration for Motion Assistant. One agent is connected to Skills, web tools, sandbox capabilities, retry and loop-guard middleware, and operational controls. The same visual model can combine multiple agents with deterministic Workflow nodes when a process needs stricter control.
- Everything is a governed plugin — install and evolve models, integrations, middleware, Skills, MCP tools and Apps, Assistant templates, Remote Components, Workbench views, and complete Agentic Apps through one lifecycle.
- Agent and Workflow hybrid architecture — use agents for flexible reasoning and workflows for stable, inspectable control paths.
- Governed enterprise execution — expose data and business actions through typed tools, semantic objects, policies, approvals, and audit trails instead of handing models raw access.
- Human-reviewable workbenches — let tool calls open focused UI views where users can inspect, correct, approve, or submit results.
Explore Xpert's first-party Agentic Apps in the official App catalog:
The Docker path requires at least 2 CPU cores, 4 GiB RAM, Docker, and Docker Compose.
git clone https://github.com/xpert-ai/xpert.git
cd xpert/docker
cp env.example .env
docker compose up -dOpen http://localhost/ and complete the initialization flow.
For deployment options, environment configuration, and upgrades, follow the self-hosting documentation. For source development, use a supported Node.js LTS release and the repository-pinned pnpm version through Corepack.
| Area | What it enables |
|---|---|
| Agent Studio | Visual authoring for digital experts, multi-agent collaboration, Workflow nodes, toolsets, knowledge, Skills, and Agent Middleware. |
| File and Knowledge Understanding | Parsed file assets, chunks, page images, citation anchors, retrieval, GraphRAG-style evidence, and workspace files. |
| Agentic BI and Data Xpert | Semantic models, indicators, natural-language analysis, and governed object-semantic tools for enterprise data queries and actions. |
| Agentic Apps and Workbench | Plugin-delivered business applications with Assistant tools, review views, Remote Components, configuration, and lifecycle hooks. |
| MCP, Skills, and Plugins | Reusable integrations, model providers, tools, middleware, managed runtimes, and installable capability packages. |
| ChatKit and Embedding | Streaming assistants for React, Vue, Angular, SAP UI5, Web Components, and vanilla JavaScript, with files, threads, tools, widgets, and i18n. |
| Operations | Conversation and task state, tool-call events, usage reporting, logs, metrics, retention controls, and runtime monitoring. |
Xpert follows an Agent-Workflow Hybrid Architecture. Agents decide how to solve open-ended tasks; workflows make critical paths repeatable and reviewable. Tools, knowledge, data resources, and plugin surfaces remain behind explicit contracts and governance boundaries.
Read the architecture article.
| Repository | Purpose |
|---|---|
xpert-plugins |
Official and community integrations, model providers, middleware, tools, Skills, and Agentic Apps. |
chatkit-js |
Embeddable ChatKit packages, widgets, and examples for multiple frontend frameworks. |
xpert-sdk-js |
TypeScript SDK packages and examples for calling Xpert APIs. |
xpert-skills |
Agent Skills for setting up Xpert and developing plugins, Agentic Apps, Assistants, and pipelines. |
docs |
Product, AI, plugin, data, BI, deployment, and tutorial documentation. |
Use the development skills from xpert-ai/xpert-skills to let a supported coding agent set up the platform and build Agentic Apps against a verified local Xpert instance.
Install the two primary skills and their plugin lifecycle companion for Codex. Remove --global if you only want to install them for the current project.
npx skills add xpert-ai/xpert-skills \
--skill xpert-platform-local-environment \
--agent codex \
--global
npx skills add xpert-ai/xpert-skills \
--skill xpert-agentic-app-developer \
--agent codex \
--global
npx skills add xpert-ai/xpert-skills \
--skill xpert-plugin-development \
--agent codex \
--globalReplace codex with another target supported by the Skills CLI when necessary. Start a new agent session after installation if the skills are not discovered immediately.
Then ask the coding agent to run the workflow in order:
-
Set up and verify Xpert
Use
$xpert-platform-local-environmentto clone or reuse an Xpert checkout at<path>, set it up in source-hybrid mode, start it, and return a plugin-test-ready environment receipt. -
Build the Agentic App
Use
$xpert-agentic-app-developerto design, implement, securely deploy, and verify<app description>as an independent Xpert Agentic App plugin against the verified local platform.
The environment skill defaults to Docker-managed infrastructure with the API and Cloud UI running from source. It verifies checkout and process provenance, service health, and whether required human initialization is complete. The Agentic App skill covers plugin architecture, Agent middleware, Workbench and extension views, Assistant templates, marketplace appConfig, deployment, and acceptance. Follow its handoff to xpert-plugin-development for plugin packaging and deployment details.
The current roadmap prioritizes:
- Project workspaces for planning, files, teams, and task execution.
- Stronger governance, approval, audit, and role-based access controls.
- Deeper trace and evaluation across Agent runs, workflows, tools, and context usage.
- Monitoring, retention, runtime controls, and production hardening for self-hosted deployments.
- Report bugs or request features in GitHub Issues.
- Read the contributing guide before opening a pull request. Contributions should target the
developbranch. - Business inquiries: [email protected]
If Xpert is useful to you, please consider giving the repository a star.
The Xpert AI Platform Community Edition is licensed under the GNU Affero General Public License v3.0. Small Business and Enterprise commercial licenses are also available. See LICENSES.md for the complete terms.