Solution · Powered by Probe

SIP Call
Testing.

A voice regression that reaches production is a customer-facing incident, a support spike and a rollback. Catch it in the pipeline instead: functional regression on every release, codec coverage, DTMF, hold, transfer and T.38, gating releases on pass/fail through GitHub Actions and the Sipfront MCP server. Ship faster because you ship safely. Powered by Probe.

What you test

Real calls, real signalling, real media.

A functional regression book exercises the call flows your customers actually use, with the same forensic capture you would get from a production trace.

Codec coverage

Validate every codec your stack should negotiate. Catch a transcoding regression before it ships.

DTMF, hold, transfer

Exercise mid-call signalling: DTMF detection, hold/resume, blind and attended transfer.

T.38 fax

Confirm fax still works after a change. T.38 negotiation and page delivery, validated end-to-end.

MOS, jitter, loss, RTT

Real call-quality metrics on every functional run, not just pass/fail on signalling.

Content validation

Confirm the right audio reaches the far end. DTMF echo, prompt detection, call-content checks.

Routing assertions

Assert that a call lands on the expected peer or destination after a routing change.

How it works

From PR to pass/fail in your pipeline.

1
Design the book
Build a regression book covering codecs, DTMF, hold, transfer and T.38. One book, every scenario.
2
Wire the trigger
GitHub Actions on each PR, or the MCP server from any pipeline. On-demand from the dashboard too.
3
Gate the release
The run returns pass/fail and full traces. Block the merge on a regression, ship on green.
4
Keep the evidence
Every call captured. When something fails, the trace is already there, no re-running to reproduce.
Scheduled vs on-demand

Beat watches. Probe gates.

Scheduled regression that runs whether or not you shipped is monitoring (Beat). Regression that runs because you shipped is release gating (Probe). Same engine, same evidence format, different trigger.

Per-PR gating (Probe)

Triggered by a deploy or a pull request. Answers “is this change safe to ship?” Blocks the merge on a regression.

Weekly cadence (Probe)

On-demand regression on a release rhythm. Run the full book before each planned release window.

Continuous watch (Beat)

Scheduled checks that run regardless of releases. Answers “is production still healthy right now?”

Gate your next release on real calls.

Try Probe free for 14 days. The AI Assistant builds your first regression book and wires it into GitHub Actions. No credit card. Resume any time if you pause.