Docs — Word Is Bond

Overview

Word Is Bond is the measurement layer for voice. You point it at a voice agent you run — on any SIP / WebRTC or PSTN endpoint — and it drives synthetic persona callers against that agent, scores the conversation, and reports regressions and drift. The deliverable on every run is a scorecard: an acceptance-test artifact you (or your client) can trust.

No real customers are ever contacted. A persona caller may only dial a target you have verified you control, or the shared sandbox agent that is pre-verified for everyone.

Quickstart

Run your first scored test call:

Create an account — the free tier needs no credit card.
Add a target: the SIP URI, WebRTC endpoint, or phone number of an agent you control (or select the shared sandbox agent).
Pick a persona and a scenario — the caller's goal and the behavior you expect.
Run the call and open the scorecard.

Or start a run directly against the API:

POST /api/tests/run
{
  "scenario": "front-desk-booking",
  "targetAgent": "sip:[email protected]",
  "persona": "first-time-caller",
  "language": "en",
  "concurrency": 1
}

# → { "runId": "4f2a…" }
GET /api/tests/4f2a…   # → 200, metrics + scorecard + transcript

The scorecard

Every run returns a four-layer scorecard with a pass/fail result and a plain rationale. The layers separate a network problem from a script problem from a missed outcome, so a red result tells you where to look.

Layer	What it measures
Infrastructure	Audio quality and latency — round-trip time, jitter, dropped audio.
Execution	Transcription accuracy (word-error rate) and whether the agent stayed on-script.
Reaction	Caller sentiment across the conversation.
Outcome	Whether the task completed — the booking, the quote, the qualification.

Each run persists its metrics, scorecard, transcript, and a recording, so any result is auditable after the fact.

Scheduled runs & regression

Point-in-time testing catches today's bug; scheduled runs catch tomorrow's drift. Save a scenario into a suite and schedule it — weekly, daily, or hourly depending on your plan. Every run is re-scored against the same rubric, so a score drop is flagged the moment a model update, platform change, or prompt edit moves the agent off its known-good behavior.

This is the difference between discovering a regression in a report and discovering it from a lost customer.

CI integration

Word Is Bond is built to run in the same pipeline that ships your agent. Gate a deploy on a suite: run it on every change to the prompt, model, or integration and fail the build if the score drops below your threshold.

# Pseudocode — run a suite and gate on the result
runId=$(curl -s -X POST https://api.wordis-bond.com/api/tests/run \
  -H "Authorization: Bearer $WORDIS_BOND_KEY" \
  -H "Content-Type: application/json" \
  -d '{"suiteId":"front-desk"}' | jq -r .runId)

curl -s https://api.wordis-bond.com/api/tests/$runId \
  -H "Authorization: Bearer $WORDIS_BOND_KEY" | jq -e '.result == "pass"'

Transports & the free tier

Two transports carry a test call. Direct (SIP / WebRTC audio) has near-zero marginal cost — that is why the free tier is generous and needs no credit card. PSTN routes over a real carrier when the phone path itself is what you are testing; it is metered per minute and available on paid plans.

The free Starter tier hard-stops at its quota with no overage and no surprise bill. See pricing for the full matrix.

The ownership boundary

Word Is Bond tests your own systems. Testing your agent involves no outbound activity to a third party — the only thing being called is a target you control or our sandbox. Before a persona caller can dial a target, you verify you own it. That boundary is the product.

Run your first test

Free tier, no credit card. Open the Studio and point it at an agent you run.

Open Studio