AI that works by your rules.

Give each agent one job and the AI model that fits it. Your rules check every move. Run it in the cloud or on your own servers.

Request accessSee how it works

Follow one change

Example: a campus chiller
  1. 1

    Something changes

    A reading comes in from the equipment you already have.

    Source
    Honeywell EBI
    Site
    East campus
    Unit
    Chiller 2
    Reading
    Supply 9.8 °C, rising
    Received 08:52
  2. 2

    The AI suggests a fix

    Peak Demand looks at the plant, power prices and the weather, then suggests a better schedule.

    Action
    Set chiller schedule
    Target
    Chiller 2
    Window
    11:00 to 14:00, 21 °C
    Reason
    Power peaks 4 to 7 pm
    Proposed
  3. 3

    Your rules check it

    Plain code, not the AI, checks every limit before anything moves.

    Comfort
    Inside 21 to 24 °C
    Rooms
    Occupied rooms untouched
    Approval
    Inside agreed limits
    Checked
  4. 4

    It happens once

    Send the same change twice? It’s sent once.

    Plan
    FB-20452
    Repeat
    Recognized, not run again
    Result
    Schedule written
    Ran once at 08:53
  5. 5

    You get a receipt

    See the checks, who approved when needed, and what changed. Undo restores this sample schedule.

    Receipt
    FB-20452
    Outcome
    Done, can be undone
    Receipt kept

Shown on sample equipment.

Build agents for any job.

Build an agent for any job, like keeping a building comfortable or catching late orders, and pick the AI model that suits it best. You set what it may read, what can run within agreed limits and what needs approval.

Order RiskGuest ComfortException RoutingOperations Analyst
Browse the agents
Reads only what it’s allowed to see
Chiller 2 supply at 9.8 °C, rising
Tariff power peaks 4 to 7 pm
Every reading says where it came from.
Site East campus
Source Honeywell EBI
Equipment Chiller 2
It suggests a plan. Your rules decide if it happens.
Plan run chiller 2, 11 am to 2 pm

How it fits together.

Your systems at the bottom, your agents and rules in the middle, your team on top. Letra explains each step as it happens.

Letra

New delivery estimate: Thursday. The receipt is in BearScope.

Sample company
Your appsWhere your team works
BearScopeOrder 48822Ridgeline Supply Co.Two-day · approved by Robin K.
FibricYour agents and rules
  1. UPSGround estimate FridayShopifyPromised Thursday
  2. Order RiskShip 48822 two-day
  3. Inside the $40 upgrade limitNot shipped yet
  4. as Robin K.
  5. Sent onceShopify · 10:05
Your systems468 connectors, 285 data feeds
UPSKustomerShopify

Each customer’s data stays separate.

Customer identity travels with the request. Access checks keep covered data inside the right workspace.

Trust and security

Every plan gets checked.

The AI writes the plan. Plain code checks it against your limits before it is sent and recognizes repeated requests.

How plans are checked

Swap anything out.

Agents ask for a job, like “open a ticket”, not a specific product. So the agent’s job can stay the same when you change a model or connect a different help desk.

Read the docs

Your cloud or your servers.

Run Fibric in the cloud, or on-premises right next to the systems it works with. Either way, you decide where your data lives.

What’s available

Start with one job.

Tell us the one thing you’d most like off your plate. We’ll set Fibric up around it, with you.

Request access