Agentic AI-ready end-to-end testing framework for enterprise contact centers. Autonomously orchestrates real voice calls, multi-browser CRM verification, and telephony API validation — all in a single declarative test scenario.
The first open-source tool that tests across telephony, CRM UI, and backend records simultaneously — no mocks, no stubs, real calls.
Click to watch: Automated Voice Testing for Contact Centers (4:37)
Traditional contact center testing is manual, slow, and siloed. Each tool covers one layer:
| Traditional Approach | Limitation |
|---|---|
| UI tools (Provar, Selenium) | Test the CRM, can't place real calls |
| Telephony tools (Twilio scripts) | Make calls, can't verify what the agent sees |
| API tools (Postman) | Query records, can't coordinate timing |
| Manual QA | Expensive, slow, non-reproducible |
Audrique is an autonomous testing agent that orchestrates all three layers in parallel — because contact center bugs live at the boundaries between systems.
| Capability | Description |
|---|---|
| Autonomous call orchestration | Places real calls via Connect CCP or Twilio REST API, navigates IVR menus with DTMF |
| AI-to-AI voice testing (NL Caller) | Gemini Live-powered AI caller simulates real customers against Agentforce with configurable persona, tone, accent, and conversation assertions |
| Parallel Agentforce testing | Simultaneous multi-provider calls to validate AI agent concurrency and supervisor monitoring |
| Transcription-driven IVR | Local whisper.cpp transcribes IVR prompts in ~2.5 s, matches keywords, sends DTMF on match |
| Multi-agent browser verification | 3 parallel Playwright sessions (Agent, CCP, Supervisor) |
| Declarative scenario DSL | JSON-based scenarios — no code to write |
| Visual Scenario Studio | Drag-and-drop wizard builds scenarios at localhost:4200 |
| Org auto-discovery | SOQL-powered discovery of queues, skills, routing, business hours |
| Video evidence capture | Parallel recording + FFmpeg merge with speed modulation and annotations |
| Enterprise security | Vault integration, zero plaintext secrets, session isolation |
| CI/CD ready | Headless execution, structured JSON results, exit codes |
| Scenario Type | What It Validates |
|---|---|
| Inbound call + agent accept | Call routes correctly, screen pop appears, VoiceCall record created |
| IVR navigation (single/multi-level) | DTMF routing, queue assignment, timeout fallback |
| Skill-based routing | Agent skills matched, PendingServiceRouting records correct |
| Supervisor observation | Queue monitoring, agent offer visibility in Command Center |
| Business hours / closed message | After-hours routing, system prompt, auto-disconnect |
| Voicemail | No-agent fallback, voicemail recorded, SF record created |
| Callback | Queue-full handling, callback task created in Salesforce |
| Real-time transcript | Speech captured, transcript panel updates live |
| Hold / resume / ACW | Full call lifecycle with disposition |
| Agentforce (AI agent) | Parallel calls from CCP + Twilio, AI greeting verification, supervisor active count |
| AI-to-AI conversation (NL Caller) | AI customer calls Agentforce, validates greeting, issue resolution, conversation quality, with local WAV recordings |
- Node.js 18+
- A Salesforce org with Service Cloud Voice enabled
- Amazon Connect instance linked to the org
git clone https://github.com/snehalsurti12/audrique.git
cd audrique
npm install
npx playwright install chromiumcp instances/default.env.example instances/myorg.env
# Edit instances/myorg.env with your Salesforce + Connect credentialsINSTANCE=myorg npm run instance:auth:sf
INSTANCE=myorg npm run instance:auth:connect# Single test
INSTANCE=myorg npm run instance:test:ui:state
# Full E2E suite
INSTANCE=myorg npm run instance:test:e2e:v2
# Dry run (validate without executing)
INSTANCE=myorg npm run instance:test:e2e:v2:drynpm run studio
# Open http://localhost:4200Visual wizard for building test scenarios — suite management, connection config, advanced settings, run from browser with live output.
- Setup Guide — Org configuration, project structure, npm scripts, CLI usage
- Scenario Reference — Step actions, assertion types, full DSL reference
- Security — Vault integration, credential management, compliance
- Video Evidence — Recording pipeline, merge modes, output format
- Changelog — Version history and release notes
Contributions welcome. Please:
- Fork the repo
- Create a feature branch
- Follow existing code patterns
- Test against a real Salesforce + Connect environment
- Submit a PR with description of changes
Built with Playwright, FFmpeg, and a lot of real phone calls.