SonetXAI engineering platform

A full engineering team, one terminal command away.

SonetX is two modules under one product. The SonetX CLI runs your pipeline, from requirements to docs, through a private gateway of specialist AI personas. SonetX Test Generation reads your requirements and API contracts and writes test suites traced back to them.

One gateway, one audit trail, and no persona prompts ever sent to the client.

A run plans itself, stops to ask a real question, resumes, and hands off from persona to persona. The test engine then turns the requirement into traced scenarios.
A run plans itself, stops to ask a real question, resumes, and hands off from persona to persona. The test engine then turns the requirement into traced scenarios.
8SDLC personas in the default setup
6 phasesin the BDD pipeline, from plan to adversarial review
5 hopsof impact-graph traversal per requirement
0persona prompts sent to the client
Module 1 / SonetX CLI

Eight personas, three stages, one orchestrator.

An orchestrator plans each run and skips stages that don't apply, so a bug fix starts at implementation rather than requirements. When a persona is blocked, the run stops and asks you a specific question instead of guessing.

STAGE 01

Plan and design

  • Requirements persona captures scope and constraints
  • Architecture and API design drafted before any code is written
  • Orchestrator delegates each stage to the right specialist
STAGE 02

Build and verify

  • Implementation persona writes the code
  • QA persona verifies the change and writes a test report
  • Code review persona flags issues with read-only tool access
  • Security persona audits before release
STAGE 03

Ship and track

  • Documentation is its own stage, not an afterthought
  • Every artifact stored durably in the pipeline folder
  • A live activity view shows which persona is working and what it costs
  • Every run ends in a structured report: done, blocked, or needs review. Nothing fails silently.
Module 2 / SonetX Test Generation

From requirement to traceable test suite.

Reads requirements, specs and API contracts, builds a knowledge graph from them, and generates BDD tests mapped back to the requirement each one proves.

STAGE 01

Ingest and analyze

  • Detects and parses OpenAPI, Swagger, Postman collections and Markdown API docs
  • Self-healing extraction splits a document recursively when it's too large for the model's context
  • Requirements and their dependencies stored as a knowledge graph, not a flat list
  • Semantic deduplication merges requirements that say the same thing
STAGE 02

Reason and generate

  • Pluggable reasoning: chain-of-thought, multi-hop retrieval or document analysis
  • A multi-phase BDD pipeline: plan, discover scenarios, write Gherkin, validate
  • Enforced variety: authentication, edge cases, non-functional limits and security scenarios
  • An adversarial critic reviews every scenario before handoff
STAGE 03

Govern and track

  • A traceability matrix links every test to its requirement, source file and coverage
  • Coverage gaps surfaced with recommended scenarios
  • When a requirement changes, see which tests and components it touches
  • Cost tracked per project, document and pipeline run
Capabilities

Run it the way your team works.

Interactive

Describe the feature. The orchestrator plans the stages, and specialists ask blocking questions you answer before the run resumes.

Headless

Start runs from CI or on a schedule, and consume a structured event stream.

MCP per persona

Attach Model Context Protocol servers to individual personas, with tool access scoped to each one.

Hard tool allowlists

Each persona gets only the tools its job needs. The report-only reviewer can't edit files or run shell commands.

Security & deployment

Built to be checked, not just trusted.

Prompts
Persona prompts stay on the gateway. The client holds only a thin, signed stub that local instructions can't override.
Access
Keys issued by your admins. No public free tier and no external sign-in.
Isolation
Test Generation scopes every query to its project. Nothing leaks across projects.
Cost
Token usage and cost logged per project, document and run, for audit and chargeback.
Inputs
OpenAPI, Swagger, Postman collections, Markdown API docs, requirements and specs.
Why it's different

What changes with SonetX.

AreaWithout SonetXWith SonetX
PipelineManual handoffs between roles and toolsOne orchestrator plans and delegates
Test coverageHand-written, drifts from requirementsGenerated from requirements, traced to source
Requirement changesImpact found late, often by accidentImpact analysis from the knowledge graph in seconds
ToolingA coding assistant and a separate test toolOne gateway, two modules, one audit trail
Next step

Run both modules against your own project.

Tell us what you're building and where testing hurts today. We'll run SonetX against it with you.

connect@siviquantlabs.com