Installation & Setup
Requirements, installation in Docker or on Windows, first sign-in, the license and four steps to the first working factory.
Installation in Docker
The compose file, every environment variable, the .env file, data storage and running without compose.
Installation on Windows
The installation wizard: a Windows service, the data folder, sign-in and the license, local servers in WSL, updates and removal.
Getting Started
First login, your first server, factory and agent — and the typical path from a bare machine to a protected home page.
Factories
The deployment unit where agents are assembled: the slug, folders and ports, shared Postgres and Git, shared variables (including the ones an agent writes from a chat), Nginx routing and certificates, the home page, central login, backups and handing a factory to a client.
Agents
The creation wizard, life cycle, fleets in the agent’s header, the manager skill for self-setup, products, the nginx subfolder address, compose.yml, extra links and protection behind the central login.
Servers
The physical machines: SSH or local WSL, one set of shared credentials per machine, connection monitoring and deletion protection while factories reference it.
Hardware
CPU, memory and disks of every machine: the list, a detail page with cores, load and charts, narrowed by the fleet in the header.
Own Agent
The built-in assistant: a sidebar block with the “Tasks” button and its counter, a chat panel, model quotas and the context bar, a prompt queue, voice input and chat labels.
Tasks and Bots
Scheduled tasks, bots, the prompt queue, MCP servers and skills — right in the agents’ chats.
Chats
Centralized chats of all agents: a paginated list with “Factory” and “Agent” columns, statistics and real-time updates, a label chosen in the title row; the same list opens as the “Chats” screen in Management.
Products
Eleven ready-made apps: MultiAgent, Research, Analytics, Data, Proxy, Chatbot, Telegram, Testing (interface testing), CRM, News and Portal — each with its own agent.
Tools
Docker service templates installed on factories: deployment, life cycle, and protecting tools behind the central login.
Tasks
The queue of long jobs across all factories: a live log, a stop button, and a week of history.
Tokens, cost and speed
Three screens about money and pace: how many tokens go, what they cost, how fast the agents work; periods, breakdowns by factory, agent, model, skill and app, distributions over hours and days, spend ceilings and the export.
Accounts
Tracking LLM accounts, quotas and balances, logging into Codex.
Nginx Routing
Each factory runs its own Nginx container: agents, products, tools, agent links and custom routes in subfolders; certificates, central login and the home page at the root; MultiAgent itself can also live in a subfolder.
Factory Home Page
A static card board at the factory root: agents, apps and tools, with “Git”, “QR login” and “Logout” buttons in the corner; the page language follows the brand.
Agent API
The machine API /api/agent/* with X-API-Key: factories, agents, tools, products, models, skills, licenses and checks.
Security & Licensing
Encryption of secrets, the machine API (Agent API), rsync backups, agent licenses.
Management
A control room for when you have many agents: eight screens — tokens, cost, speed, hardware, agents, chats, decisions and incidents; fleets and their chips, down agents, seven verifiable agent states, a decision queue, spend ceilings and autonomy policies.
Environments
A library of named environment variable sets: create them, import from a factory or an agent, push to the agents and factories you choose.
Factory Handover
How to hand a factory over to its own MultiAgent: preview, a one-time operation, logging into the client’s instance without a password.
Analytics and exports
The same numbers for external systems: metrics for Prometheus, the report over GraphQL and the long-term archive in ClickHouse.