Skip to main content

How OtoDock works

OtoDock is one platform you run on your own infrastructure. A handful of pieces work together to turn AI subscriptions into a team of capable agents you drive from a browser. This page is the mental model behind everything else in the docs.

The pieces​

  • The server is the heart of OtoDock. It runs your agents, holds their files and memory, talks to AI engines, manages tools, schedules work, and serves the dashboard. Everything lives here, on your hardware.
  • The dashboard is the web app you use OtoDock through. It connects to the server over a live link, so an agent's work — text, reasoning, tool calls, file edits, plans, and to-do lists — streams to your screen as it happens.
  • Agents are the workers. Each one has its own persona, files, tools, memory, and settings. You chat with them, hand them tasks, put them on a schedule, and get notified when they have something for you.
  • AI engines are what actually power an agent's thinking. You connect your own subscriptions and keys; OtoDock runs each agent on the engine you choose.
  • Tools (MCPs) give agents real abilities — reading and writing files, browsing the web, talking to GitHub or Notion, generating images, and much more.
        Your browser
             │
        ┌────┴────┐  live connection
        │Dashboard│
        └────┬────┘
             │
     ┌───────┴────────┐
     │  OtoDock server│ ── runs agents, files, memory, schedules
     └───────┬────────┘
             │
   ┌─────────┼──────────────┐
   │         │              │
AI engines  Tools (MCPs)  Your data
(your subs) (capabilities)  (on your hardware)

Agents​

An agent is a self-contained worker: a name, a personality, a set of tools, its own folder of files, and long-term memory. You might run a personal assistant, a coding agent, a monitoring agent, and a research agent side by side — each configured for its job, each with access to exactly what it needs.

A fresh install comes with one ready-to-use assistant. You add more by installing them from the community catalog or building your own. Agents can also work together — several can join the same meeting to tackle a problem from different angles.

→ Learn more in Agents.

Sessions​

When you chat with an agent, OtoDock starts a session for it — a live, running instance of that agent with your conversation loaded. Sessions are what make agents feel responsive: the agent stays warm between your messages, so follow-ups are fast and it keeps full context of what you've been discussing.

You don't manage sessions directly — OtoDock does it for you:

  • Warm-up. The first message spins the agent up and loads its tools. Subsequent messages reuse the running session.
  • Resume. Conversations pick up where they left off, even after the server restarts. Your history is saved, so nothing is lost.
  • Idle clean-up. A session that sits unused is quietly retired to free up resources. The next message simply starts it again — your conversation is still there.

The same machinery powers chats, scheduled tasks, and meetings, so an agent behaves consistently no matter how it's working.

AI engines​

OtoDock doesn't ship a model — you bring your own. The real work runs on Claude Code and Codex, the two most capable coding agents available, using subscriptions and keys you already have. A lightweight direct engine is also available for low-latency cases like voice.

You pick an engine and model per agent, and can switch on the fly from any chat.

→ Learn more in AI Engines.

Tools​

On their own, AI models can only produce text. Tools — built on the open Model Context Protocol (MCP) — are what let an agent do things: edit a document, run a command, search the web, open a pull request, write up a page in Notion. OtoDock ships with a solid set of built-in tools, integrates with popular services, and lets you install more from the community catalog or build your own.

→ Learn more in Using Tools (MCPs).

Where agents run​

Every agent runs inside a strict sandbox on the OtoDock server: its files and network access are locked down, and you decide exactly what each agent can reach. This is the safe default for running AI on infrastructure that holds your data.

→ Learn more in Security.

Your data stays yours​

Everything — conversations, files, memory, credentials — lives on the server you run. AI subscriptions and API keys are stored encrypted, and a running session carries only the credentials its own work needs — never anyone else's. OtoDock talks to the AI providers you connect and to the services you explicitly integrate, and nothing else.

Next steps​

  • Agents → — personas, roles, and how agents collaborate.
  • AI Engines → — Claude Code, Codex, and the direct engine.
  • Security → — the sandbox and what agents can and can't reach.
  • Features → — chat, tasks, triggers, meetings, voice, and more.