IN Brief:
- VIAVI has integrated a Model Context Protocol server framework into its NITRO AI test portfolio.
- TestCenter, CyberFlood, and TeraVM expose different network and security test functions to compatible AI clients.
- The framework includes automated configuration correction and human review checkpoints for critical operations.
VIAVI Solutions has extended its NITRO AI testing portfolio with a Model Context Protocol server framework that allows compatible AI clients to invoke functions on network and security test platforms. The integration covers VIAVI TestCenter, CyberFlood, and TeraVM, providing an interface between natural-language instructions and the configured instruments or virtual test environments used for validation work.
TestCenter addresses network performance under controlled traffic, while CyberFlood covers security and application performance and TeraVM handles virtualised testing, including secure access and zero-trust network access. A single validation programme may use more than one of these environments, so connecting their supported functions to an agent interface creates a common orchestration route across otherwise different software and test platforms.
Model Context Protocol provides a standard interface through which a client can discover tools exposed by a server and request their execution. In VIAVI’s implementation, compatible AI clients can select supported test operations, configure them, and retrieve outputs. TestCenter, CyberFlood, and TeraVM still generate the measurements or simulated traffic; the new layer changes how those functions are called and coordinated.
A network validation sequence begins with defined traffic profiles, port and topology settings, and the measurements required for acceptance. Throughput, latency, error rates, and application behaviour become comparable only when the configuration is consistent between runs. The MCP layer can assemble the available operations into a workflow, while the underlying test tools execute it against the chosen hardware or virtual service.
Existing laboratories already use vendor APIs and scripts to repeat those sequences, but dependencies and parameter mappings have to be maintained as configurations evolve. Exposing supported functions through the MCP server gives compatible AI clients a structured interface to that same infrastructure. Scripted automation remains available; the additional control method is designed to reduce the manual steps involved in constructing and coordinating workflows.
The company has also incorporated an automated error correction loop intended to detect test-tool configuration problems and attempt to correct them. Misconfigured ports, inconsistent test parameters, and incompatible arrangements can prevent a result from being meaningful even when the test executes. A software-assisted configuration loop may reduce routine correction work, but the validity of the result still depends on whether the test conditions match the system’s actual design and acceptance requirements.
Critical actions can remain behind human review checkpoints while lower-risk configuration and reporting steps proceed automatically. That control is particularly relevant where a test will affect shared ports, generate substantial traffic, or change an environment used by another team. Each workflow also needs an unambiguous account of configured conditions and returned results so that failed runs can be reproduced.
Once results are available, the framework is designed to support workflows that respond to them: a configuration error can trigger an attempted correction and a rerun, while a completed preliminary test can lead to the next defined measurement. This feedback loop builds on established conditional test automation rather than changing measurement physics. Its reliability depends on the permitted operations, the supplied test parameters, and appropriate review of the resulting outputs.
Network hardware development now commonly includes Ethernet fabrics, high-speed interfaces, virtualised network functions, and security components that must operate together. A test plan can span traffic generation, application delivery, and secure-access behaviour even when the equipment under examination is a single integrated system. The named VIAVI platforms address those areas from different directions, making a common control method potentially useful for multi-platform laboratory work.
AI-initiated execution must still use controlled workloads, known thresholds, and repeatable conditions if results are to inform qualification or fault isolation. A natural-language instruction that omits traffic profile, topology, or pass criteria requires explicit configuration choices; the same ambiguity would affect a hand-written script. Keeping engineering acceptance rules separate from the assistant interface is necessary for comparing tests over time.
The NITRO AI framework now exposes an agent-facing control path across TestCenter, CyberFlood, and TeraVM, with configuration correction and review checkpoints in the workflow. Its practical use will be determined by how laboratories map those callable functions to established test procedures, access controls, and evidence records. VIAVI’s development changes the control route for testing while leaving the underlying instruments, traffic models, and acceptance requirements in place.
