Introduction
A two-part platform — a hosted control plane and a runtime binary you run yourself.
Welcome to the docs. The platform is made of two parts that work together:
- The control plane (this dashboard) — sign in with W3 ID, create projects and agents, upload workspace files, and manage how LLM requests are routed. It also acts as an authenticated proxy for LLM calls, so model API keys never leave the control plane.
- The runtime (
bot-agent) — a standalone binary you download and run yourself on your own machine, server, or container. It handles conversations, tool execution (shell, browser, cron), scheduling, and persistence. Your application talks to it over a local HTTP and WebSocket API.
Your application
│
├── calls ──→ bot-agent (you host it, you run it)
│ │
│ └── calls ──→ control plane /api/runtime/*
│ (LLM proxy + workspace)
│
└── manages ──→ dashboard (projects, agents, files, users)
How the pieces fit
| You do… | Where it happens |
|---|---|
| Create projects, agents, members | Dashboard |
Upload workspace files / AGENTS.md | Dashboard (project files) |
| Download the runtime binary | Dashboard (agent page) |
| Run the agent process | Your infrastructure |
| Send messages, stream responses | Your app → bot-agent |
| Route LLM calls to vLLM backends | Control plane (automatic) |
The control plane never starts, stops, or proxies to the runtime. You own the runtime process entirely. It calls back to the control plane only for LLM access and workspace content.
Where to go next
| If you want to… | Read this |
|---|---|
| Ship your first agent | Getting started |
| Understand the runtime model | Architecture |
| Run the binary in production | Deployment |
| Keep tokens & keys safe | Security |
| Manage reusable credentials | Secrets & Vaults |
| How model routing works | Models & routing |
| Integrate with the HTTP API | Agent API: Getting started |