xorbitsai/xagent
Enterprise agent platform for building, deploying, and operating AI teammates.
What it solves
Organizations that put AI agents in front of real users need control: what the agent can access, where users reach it, and how to investigate a bad result. Xagent gives you a workspace to run tasks directly and a way to publish and operate reusable "teammates" — agents configured for specific recurring jobs — with visibility into every request.
How it works
You describe a task in the workspace and run it right away; in Auto mode Xagent can answer directly, call tools, or plan and execute multiple steps. For repeatable work, you build a teammate in the Builder: give it a job description, pick a model, and attach skills, knowledge bases, and MCP connectors for external tools. Then publish the teammate and choose an entry point — a shareable link, an embedded widget on allowed domains, or a REST API with an agent API key. After launch, Conversation Logs let you find conversations by teammate or source and inspect transcripts, execution steps, and tool calls when trace data is available. Teammates can also run on schedules or respond to webhooks, and multiple specialists can be combined into a Workforce managed by a manager agent. You can self-host with Docker Compose, connect supported model providers or serve your own via Xinference, and enable sandboxing for eligible tool calls.
Who it’s for
Teams and enterprises that want to deploy AI agents for internal or customer-facing work with proper governance — controlling access, choosing where users interact, and auditing results. It suits both those who want a hosted option (TobyBot Cloud) and those who need to self-host and manage their own infrastructure and user accounts.
Highlights
- Run tasks directly from the workspace with Auto mode for multi-step planning and tool use
- Build reusable teammates with configurable instructions, model, skills, tools, and knowledge
- Publish teammates via shareable link, embedded widget, or REST API
- Inspect conversation logs and execution traces, including tool calls and knowledge searches
- Self-hostable with Docker Compose, or use the hosted TobyBot Cloud
- Support for schedules, webhooks, and Workforces that delegate work to multiple specialists
- Sandboxing for eligible tool calls and monitoring of model and tool usage
Related
- Project
- Project
- Project
- Project
- Project