Agentic infrastructure
The working surface beneath the views — the tool fabric, the agent roster, skills, routines, memory, the knowledge graph, the daemons, and the doctrine that governs them.
Everything you see in Command Center's views sits on a working surface you also own: the tools agents call, the agent definitions they run as, the skills and routines that package repeatable work, the memory and knowledge layers they read and write, the background services that keep it all running, and the written doctrine that makes a freshly started agent behave like a member of your operation. That surface is what this section documents — agentic infrastructure means the parts your agents operate on, not the parts you click on.
Two facts shape every page here:
- Dual surface, one system. Every capability has a human surface (a view) and an agent surface (a tool or local API) over the same data. When this section documents a tool, the matching view is documented under Features; both doors open onto the same room.
- It is all on your disk. The agent definitions, skills, routines, protocols, and memory files are plain files under your install root. The services are local processes on loopback addresses. Nothing in this section describes a hosted component — depth here is depth into your own machine.
The map
| Area | What it covers |
|---|---|
| MCP servers | The tool fabric — 16 local servers your agents call tools through, documented one page per server, tool by tool |
| Agents | The agent roster — each role, what dispatches it, and which pipeline it belongs to |
| Skills | The skill catalog — packaged expertise agents load on demand |
| Routines | The automation catalog — scheduled definitions the routines engine executes |
| Memory | The memory layers at operating depth — what persists where, and how you curate it |
| Knowledge graph | The graph at operating depth — workspaces, depths, ingest, enrichment, sync, lint, export |
| Daemons | The background services — roster, ports, lifecycle, autostart, and the loopback-only posture |
| Doctrine | The shipped operating rules — what each protocol governs and where it lives on your disk |
How the areas fit together
An agent session starts by reading the doctrine — the operating contract and the protocols it points to. It reaches every capability through the MCP servers, which mostly front the daemons doing the actual work. It recalls what previous sessions learned through the memory layers and navigates your documents through the knowledge graph. Specialist work is delegated to agents from the roster, sharpened by skills, and the recurring portion of all of it runs unattended as routines.
If you are new to the platform, read the concept pages first — they explain why the system is shaped this way (the operator model, memory layers, the knowledge graph). The pages in this section are the operating manual: the depth you and your agents work from day to day.