Automation
Rules first, a reasoner on the exceptions
Automation that proposes, and a person who decides.
From a thermostat schedule to a fleet of autonomous machines, Furcate automates through one loop: monitor, analyse, plan, act, with policy on every step and a person's approval where it matters. Rules handle the routine; a model reasons only about the exceptions; nothing acts without leaving a record.
01
One loop for everything
Edge to autonomous.
The same control loop runs a room, a plant and a fleet.
Automation reads the same signed readings and writes through the same signed orders as a person does, so automating something never opens a path around policy.
- 01
Monitor
Signed readings from every device, with their rung.
- 02
Analyse
Rules first; forecasting and a reasoning model only on what the rules cannot settle.
- 03
Plan
A proposed action on a capability, checked against interlocks and safe ranges.
- 04
Act
A signed order, approved by a person's passkey when policy requires one, enforced at the leaf or node.
- 05
Record
The ledger holds the proposal, the approval and the outcome, including every refusal.
02
Agents with limits
Bounded autonomy, delegated by people.
Agents can see the whole site and act on almost none of it alone.
Agents reach Furcate through MCP. They read what policy lets them read, propose actions, and act only within limits a person delegated; anything beyond waits for approval. Schedules, workflows and webhooks run under the same rules.