Docs
Product · Agents & Actions

From words to facts, with your approval.

Seeing is not enough. On the cases you have validated together, Meddle's agents write the new setpoints straight onto the machines. With permissions, full traceability and an override always available.

Actions · this week3 writes
GP800 · setpoint 62 → 58 °Capproved
Pump 2 · threshold 4.2 → 3.9 barapproved
Boarding-area AHU · manual modeapproved
Writes without approval0
“Who wrote a setpoint this week?”
100%
of writes logged, with the value before and after
0
writes without a human approval
Override
always available, at any moment
What it does

First it suggests, then it executes.

It flags

The agent proposes the intervention the moment the data justifies it: a setpoint to correct, a load to shift, a work order to open.

It waits for validation

During the pilot the system is read only. Writing to the PLCs is enabled only after you have validated the cases, one by one.

It writes and records

On validated cases the agent writes the setpoint to the machine. User, previous value, new value and time all land in the audit trail.

How it works

Detect, validate, act.

Detect

The condition arrives from the machine the moment it happens, not at the end of the shift.

Validate

The operator confirms. Until a case is validated the agent stops at the suggestion.

Act

The setpoint is written back to the machine, with role-based permissions and an immediate override.

The limits, by design

Where the agent stops.

Human in the loop, always. No parameter is written without a person approving it. It is not a mode you can switch off.

Regulatory constraints win. Safety and quality take absolute priority over any optimisation. The hierarchy stays safety, quality, then efficiency.

Reversible by construction. Every write is recorded with its previous value, so going back is a traced operation like any other.

Frequently asked questions

The rest, in short.

Can Meddle really change machine parameters?

Yes, but first it is read only and suggestions. Then, with your approval, the agents write setpoints onto the machines, with full permissions and traceability.

Can automation put compliance at risk?

No. Regulatory constraints take absolute priority over any optimisation, and every action is validated by a person.

How do I see what was written?

In the audit trail: every write records the user, the previous value, the new value and the time.

Can I disable writing?

Yes. Writing is enabled per case and per role, and the manual override stays available at any moment.

Want to see one action, end to end?

We walk the whole path: from the data to the suggestion, from the approval to the logged write.

Book a demo