Create your first agent
Agents are the heart of OtoDock: a name, a role prompt, an AI engine, and the tools you grant. Here's how to set one up well.
5 min read
An OtoDock agent is a persistent coworker, not a one-off chat: it keeps its own identity, memory, tools, and workspace across sessions, and everyone on your install can work with it.
1. Create the agent
As an admin, open the agents section of the dashboard and create a new agent. Give it a clear name and role — "IT Assistant", "Research Analyst", "Ops Bot" — because the name is how your team will think about it.
2. Write the system prompt
The system prompt defines who the agent is and how it works. The strongest prompts state the agent's job, its boundaries, and the conventions it should follow — short and concrete beats long and vague. You can refine it any time as you learn what the agent needs.
3. Pick the engine and model
Choose which AI engine powers the agent — Claude, Codex, your own API keys, or local models — and a default model. Different agents can run on different engines on the same install.
4. Grant tools
Tools come from MCP servers you enable per agent: files and documents, GitHub, notifications, scheduled tasks, and the growing community catalog. Start small and add tools as real work demands them — every grant is a trust decision.
Then open a chat and put it to work. The docs cover agent memory, shared workspaces, and multi-agent meetings when you're ready for more.
Put your agents to work
Self-host free up to 5 users — running in minutes.