SIPforge

How to use

How to run your first voice test.

A public walkthrough of the SIPforge workflow: pick an engine, keep the first run small, and read the result without treating every unanswered call as a failure.

The workflow

  1. Step 1

    Choose the engine

    Use SIP for your own servers and SBCs, Twilio Voice for carrier or Programmable Voice calls, and AI Voice Agent when the thing under test is a live conversation.

  2. Step 2

    Prepare the dependencies

    Add the SIP server, Twilio account and caller number, media file, voice-agent persona, or AI provider keys the scenario needs before you launch.

  3. Step 3

    Keep the first run small

    Start with one call or a very low rate, a short duration, and a cost cap on anything that bills. Raise load only after that baseline looks right.

  4. Step 4

    Watch the run while it moves

    The live view shows active calls, terminal statuses, answer seizure ratio, failures, spend, voice quality, and agent latency before the test finishes.

  5. Step 5

    Open the evidence

    Call detail is where SIP responses, Twilio events, DTMF, transcripts, timing, and per-call cost explain a number that looks wrong in the summary.

  6. Step 6

    Share it or run it again

    Export PDF, CSV, or JSON, compare a later run, or save the configuration as a template so the next test starts from the same setup.

First run by engine

SIP

Signal and media traffic against Asterisk, FreeSWITCH, Kamailio, SBCs, and gateways.

  1. 1.Create a server with host, port, transport, and credentials.
  2. 2.Confirm the firewall, NAT, and RTP range can reach the target.
  3. 3.Run OPTIONS or INVITE/BYE with one call before you raise CPS.
  4. 4.Read SIP status first, then MOS, jitter, and packet loss when the scenario carries RTP.

Twilio Voice

Real Programmable Voice calls with status, DTMF, cost, and callback evidence.

  1. 1.Connect the Twilio account and choose a caller number.
  2. 2.Call one known test number and set a low timeout and cost cap.
  3. 3.Use play-and-gather or send-DTMF only after a plain call completes.
  4. 4.Treat busy and no-answer as unanswered calls. Treat a failed status as the failure.

AI Voice Agent

A real phone conversation scored for transcript, latency, and goal completion.

  1. 1.Write a short persona, opening line, goal, and evaluation criteria you can prove from the transcript.
  2. 2.Connect telephony plus the language model, speech-to-text, and text-to-speech providers.
  3. 3.Place one outbound call with a 60 second maximum duration.
  4. 4.Read the transcript and the latency breakdown before you add concurrency.

Call status rules

Destination availability is not the same as a platform failure. Only a failed status should fail the test.

StatusMeaningEffect
completedThe call connected and finished.Counts as answered. Included in ASR.
busyThe destination was reachable but busy.Lowers ASR. Does not fail the test.
no-answerThe destination did not answer in time.Lowers ASR. Does not fail the test.
canceledThe call was canceled before it completed.Lowers ASR. Does not fail the test.
failedSIP, Twilio, or the runner reported a real failure.Counts as failed and can fail the test.

How to read the metrics

ASR
Completed calls divided by total attempts. Busy, no-answer, and canceled lower it without failing the run.
MOS
Estimated voice quality for SIP tests that capture RTP. Higher is clearer audio.
Jitter
Variation in packet arrival time. Lower usually sounds smoother.
Packet loss
Media packets that never arrived. Rising loss damages intelligibility.
Latency
Time from callee speech to agent audio on an AI voice-agent call.
Cost
Twilio telephony spend, plus AI usage when that engine is in the test.

More answers are in the FAQ. After you sign in, the workspace guide walks through each module, including templates, schedules, alerts, and exports.

Ready when you are

Start with one call, then read the result.

Create a workspace when you are ready to run the workflow on your own SIP server, Twilio account, or voice agent.

Start testing Talk to our team