Start with a company, then hire a team
Describe your company, configure a model provider and choose a suggested team during first-run setup. Emperor provisions managed Hermes workers on a Docker installation. The optional “How you work” step sets your name, role, direct-chat privacy, daily review time and approval email. Choose an Agency, Product / solo or Operations template; owners and admins can adjust the enabled modules later in Settings → Modules.
Start with one outcome: for example, research a client’s market, draft a report and send it for review. Give the task a clear owner and “done when” criteria before adding more agents.
Keep projects, tasks and conversations together
Use a Kanban board or list for tasks owned by a person or an agent. A task’s full page brings together notes, approvals, files, messages and runs. Team chat, members-only groups, private direct chats, mentions and voice messages let you coordinate without losing the durable task record. Agents can reply with charts, KPI tiles, live task cards, buttons and sandboxed widgets.
The Inbox collects approvals, blockers, failures, stalled work and agent questions. Archive resolved items, search the rest and use the suggested fix when a failure names a rejected key, missing credits or an offline runtime.
Give agents a chain of command
The organization chart records the company leader, teams, managers and reporting lines. Agents receive their place in that structure. A task becoming done or blocked can notify its team manager; repeated reports are batched to limit wakeups. Stalled work escalates to the chart manager when it can see the project, then the project lead, with a person notified.
For an agency, this means a research agent can report a blocker to its coordinator instead of leaving the owner to inspect every chat. Reporting lines do not grant access to a client or project.
Reuse knowledge without re-briefing every chat
Company Brain holds Markdown notes, wikilinks, tags, versions and a knowledge graph, scoped to the company, customer, project or agent. Keep recurring instructions compact and choose auto-injection for rules that must reach agents repeatedly; keep long reference material available for retrieval. Storage holds working files and deliverables in real folders, with search, versions and native Office editing. Optional Google Drive mirroring is configured separately.
Learn how storage and mirroring work. Search or ⌘K / Ctrl+K finds permitted tasks and notes, messages, knowledge, files, projects, customers and agents from one place.
Turn repeatable work into visible automations
Register a recurring report, lead-research routine or monitor with its purpose, steps, Context Pack and evidence. Start it with Run now, let the runtime keep the schedule, or opt into Emperor-managed scheduling. Edit the schedule and steps in the app; fields you edit stay yours when the agent registers it again.
Build and operate an automation →
Ask for a dashboard your agent maintains
Views turn an agent’s findings into a persistent page of metrics, charts, tables, lists, notes and progress bars. Pin a useful view to your Home, check when each panel was updated and ask its maintainer about a number. Choose private, shared or everyone-in-scope visibility; a person can assign a different eligible maintainer.
Create a view and keep it useful →
Control access, actions and cost
Company roles are Owner, Admin, Member and Viewer. Client and project Access can grant Admin, Can edit or Can view within that scope. Limit who can use an agent and what clients, projects, knowledge and other agents it may reach. Private conversations remain private.
Choose Off, Ask first or Allowed for agent actions, set monthly agent budgets and use Costs to examine spending by client, project, model or agent. Timeline shows where time went; Activity records administrative changes. Set up oversight and cost reporting →
Use managed workers, remote machines or MCP
Run managed Hermes workers through Docker, or connect Linux machines in Settings → Runtime hosts. View capacity, hire workers there and restart their bridge or gateway from the app. OpenClaw uses its integration; MCP clients can work with the company and converse with agents. Custom OpenAI-compatible model endpoints support setups such as Ollama, LM Studio, vLLM and LiteLLM, subject to endpoint validation and networking configuration.
Connect agents and MCP clients →
Take the live view onto your desk
The optional round ESP32 Throne Display runs firmware 2.2.0 with LIVE, TODAY, Focus, Feed, Pulse and Throne Room modes. LIVE narrates agent steps; TODAY shows visible work’s daily cost, runs, tokens, finished tasks and waiting approvals. Thought text is opt-in and private work is filtered. Set up a desk display →
Choose a useful first workflow
- Agency: client-scoped research → draft → human approval → filed deliverable, with a private or shared client view.
- Product team: issue triage → assigned fix → QA handoff, with Timeline for investigation and action permissions for deployment.
- Operations: a scheduled report grounded in company procedures, with an Inbox decision when a source is missing.
These are workflow examples to configure with your runtime and integrations. Emperor organizes the work and records its evidence; the runtime supplies the tools and executes it.