Seranai Dev Tools · In development
Find the failure. Keep the test.
Test your voice agent locally, understand what failed, and keep the regression test. Add a paid dashboard when you need deeper coverage, shared history, and team release checks.
What broke? Where? Can we catch it next time?
Changing a prompt, model, voice, or tool can change how your agent behaves. Celerto is being built to turn those changes into evidence you can inspect: the conversation, the timing, the tool calls, and the difference from your last run.
Free local testing
A complete debugging loop on your machine is the foundation of Celerto. The planned local runner and reports are free; you bring your own provider keys.
CLI and local browser interface
Run saved scenarios from your terminal and inspect results in a browser on your machine.
Reusable scenarios and recordings
Keep input fixtures and test cases with your project so failures can be exercised again.
Audio, transcript, and event timeline
Review the conversation alongside available agent events to understand what happened.
Timing and interruption metrics
Inspect response latency, interruption stop delay, overlapping speech, and long silences where instrumentation supports them.
Tool and outcome checks
Assert tool arguments and check sandbox state, such as whether an appointment was actually booked.
Compare and export
Compare with a previous run and export local reports for your development or CI workflow.
Paid dashboard: broader testing, shared evidence
When your local tests become a release process, the planned paid platform adds hosted execution and a shared dashboard. Use it to test more conditions, compare versions, and coordinate decisions across your team.
Parallel hosted runs
Run broader suites without tying up your development machine.
More conversation conditions
Exercise different scenarios and selected audio conditions, such as noise and interruptions.
Shared history and comparisons
Review results across agent versions, releases, and client projects.
Deeper evaluation
Combine measured timings, verified outcomes, and AI-assisted conversation grading with visible evidence.
Schedules and release checks
Run suites on a schedule and apply agreed acceptance criteria before releasing changes.
Production failures as tests
Bring a captured failure into the scenario library to protect future releases.
Team reports
Share results and review changes with the people responsible for shipping the agent.
Retention and permissions
Agree data retention, team access, and deployment needs for the selected plan.
Scores you can investigate
Measured
Timings, silence, interruptions, tool latency, and errors from the available recordings and events.
Verified
Tool arguments, expected backend state, required steps, and prohibited actions checked against assertions.
Judged
Helpfulness, grounding, and conversation quality evaluated with AI assistance and reviewable reasoning.
Available metrics depend on your integration and instrumentation. AI-assisted grades need calibration; an overall score does not replace inspection of the evidence.
Local means your runner and records
Recordings, traces, and results are intended to stay on your machine by default. Calls to external agent, speech, simulation, or grading providers can still transmit data and incur their own charges. Hosted testing requires an explicit upload and an agreed retention policy.
One testing workflow as you grow
We are designing the local and hosted products around the same scenarios and metric definitions. Tell us whether you build with LiveKit, Pipecat, or a custom stack; adapter support will be confirmed for each pilot.
Free locally. Paid when you need the platform.
Our planned model is a free local runner, a paid dashboard subscription, and separately explained hosted-testing usage. Paid prices and allowances are not yet finalized. There is no unlimited-compute promise.
Build with us
Celerto is in development. Share your stack, a recent failed call, and how you test changes today. We will discuss whether the early-access or dashboard pilot is a fit.
Questions and answers
What is the difference between local testing and the paid dashboard?
The planned free runner supports your local debugging loop. The paid dashboard adds hosted runs, shared history, deeper coverage, and team release checks.
Is Celerto available today?
Celerto is in development. Contact the team to discuss early access or a paid dashboard pilot.