Simulations
View transcripts, scores, audio, and AI insights from every test conversation
Simulations
A simulation is one test conversation — a specific persona talking to a specific agent, graded by a specific scorecard. It's the atomic unit of testing in Chanl. Every simulation produces a transcript, a score, audio, and AI-powered analysis of what went right and wrong.
What You Get
| Output | Description |
|---|---|
| Transcript | Full conversation text with speaker turns and timestamps |
| Score | Pass/fail + per-criteria breakdown from the scorecard |
| Audio | Recording of the simulated call |
| Analysis | AI insights — what the agent did well, where it failed, and why |
| Tool Calls | Every MCP tool the agent invoked during the conversation |
Viewing Simulations
Go to Test → Simulations in the dashboard, or navigate from a scenario's results page.
Each simulation shows the conversation timeline — who said what, when tools were called, and how the scorecard evaluated each criterion.
Via API
# List simulations for a scenario
curl "https://api.chanl.ai/api/v1/simulations?scenarioId=<id>" \
-H "Authorization: Bearer <access_token>"
# Get a specific simulation with full details
curl "https://api.chanl.ai/api/v1/simulations/<id>" \
-H "Authorization: Bearer <access_token>"The response includes the transcript, score, and metadata:
{
"data": {
"id": "sim_abc123",
"scenarioId": "scenario_xyz",
"personaId": "persona_karen",
"agentId": "agent_v2",
"status": "completed",
"score": {
"overall": 72,
"pass": false,
"criteria": {
"accuracy": { "score": 4, "max": 5, "pass": true },
"compliance": { "score": 3, "max": 5, "pass": true },
"empathy": { "score": 2, "max": 5, "pass": false },
"resolution": { "score": 5, "max": 5, "pass": true }
}
},
"duration": 124,
"createdAt": "2026-03-15T09:00:00Z"
}
}Transcript
curl "https://api.chanl.ai/api/v1/simulations/<id>/transcription" \
-H "Authorization: Bearer <access_token>"Audio
curl "https://api.chanl.ai/api/v1/simulations/<id>/audio" \
-H "Authorization: Bearer <access_token>" \
--output simulation.mp3Comparing Results
The real power of simulations is comparison. Run the same scenario with different agent configurations and compare:
| Agent | Karen (frustrated) | Priya (polite) | Maria (confused) |
|---|---|---|---|
| Agent V1 | 62 — failed empathy | 88 — passed | 71 — failed clarity |
| Agent V2 | 78 — improved | 91 — passed | 85 — improved |
This tells you exactly whether V2 is an improvement and where it still needs work.
Compare simulations across agent versions to catch regressions. If V2 scores higher on "frustrated customer" but lower on "analytical customer," you'll know before deploying.
Simulation Lifecycle
- Created — scenario executed, simulation queued
- Running — conversation in progress between persona and agent
- Completed — conversation finished, transcript available
- Scored — scorecard evaluation complete, pass/fail determined
- Failed — agent unreachable, timeout, or error