Self-Hosting
Xagent is built to run on your own infrastructure. You control the models, the data, and where everything runs — local, private cloud, or on-premise.
Deployment Options
- Local — Docker Compose on a single machine for development or small teams.
- Private cloud — your VPC, behind your own networking and identity.
- On-premise — fully air-gapped enterprise infrastructure.
Architecture
A deployment is composed of the application backend, a frontend, a database, the agent execution sandbox, and (optionally) the SaaS layer for teams and billing. An nginx reverse proxy fronts the services.
| Layer | Responsibility |
|---|---|
| Agent definition | Intent & constraints |
| Planning engine | Dynamic decomposition |
| Execution runtime | Orchestration & sandboxing |
| Tool layer | Integrations & actions |
| Model layer | LLM & inference backend |
Security
Agent code executes inside a VM-level sandbox isolated from the host, so untrusted, model-generated actions cannot reach your infrastructure directly.
Models
Bring your own model providers — OpenAI, Anthropic — or serve open-source models with Xinference for full control over inference.