Skip to content

Getting Started

Two ways to adopt IoC

There are two ways into the Internet of Cognition, and they answer different questions. Run puts IoC into a multi-agent system you operate. Test measures what it changes, before or alongside that.

Neither path requires the other. A team that wants the benefit in production can start with Run; a team that needs evidence of the benefits first can start with Test.

  • Run

    Instrument your multi-agent system and insert the Cognition Protocols, so the system handles the cognitive challenges it meets and you can see what it did.

    Go to Run

  • Test

    Reproduce the existing labs, or build your own, to measure what cognitive failures cost and what the Cognition Protocols are worth on your own system.

    Go to Test

Run

Cognition Protocols + SCOPE

Running IoC means two things, and they are independent — either is useful on its own, and together they close the loop.

Insert the Cognition Protocols into your agents, so that when a cognitive challenge arises the system has a mechanism to deal with it: ACCORD establishes shared intent before the work starts, and CONCORD reaches a decision when the agents pull apart. Both ship as skills, so nothing in the application has to be rewritten.

Instrument your system for SCOPE, which recomposes the per-agent traces into one multi-agent trajectory and gives you monitoring and explainability over the system as a whole — including the cognitive failures that latency and error-rate monitoring structurally cannot see.

Component What it adds Start here
Cognition Protocols ACCORD and CONCORD, deployed as skills in the agents or in the AgentBridge Cognition Protocols — Installation
SCOPE Instrumentation, trajectory recomposition, detection of cognitive failures, analysis and explainability SCOPE — Getting Started

Done together, SCOPE is what tells you a cognitive failure happened and the protocols are what act on it.

Test

MAS-Lab, with the Cognition Protocols and SCOPE inside it

Testing means running the system under MAS-Lab, which integrates both the Cognition Protocols and SCOPE so that a claim about either can be measured rather than asserted.

You can reproduce the existing labs — the cognitive-failure studies and the Cognition Protocols evaluation — and you can write new ones, against either the sample multi-agent systems MAS-Lab ships or your own system.

What that buys you is a number instead of an argument: what a given cognitive failure actually costs on a system like yours, and what the Cognition Protocols are worth against a matched baseline, with SCOPE's sensing layer embedded to supply the evidence.

What you want Start here
Install MAS-Lab and run a first example, then reproduce or write labs MAS-Lab

Which to start with

If you… Start with
operate a multi-agent system and want the benefit in production Run
have to justify adoption with numbers before changing production Test
are deciding which cognitive failures matter on your own system Test
already run OpenTelemetry instrumentation Run, beginning with SCOPE
own the agents and want coordination benefit first Run, beginning with the protocols

The two are not exclusive, and most teams end up with both: Test to establish what a change is worth, Run to put it in front of real traffic.