Measure what matters, then make it better.
Shipping the agent is the start. Symphia gives you the console to see how it's really performing, a test suite to prove changes are safe, and an audit trail so every improvement is traceable.
The toolkit
See it, test it, trust the change
Everything you need to turn a live agent into a continuously improving one.
Insights console
KPIs on volume, resolution, and deflection in one place — see what customers ask and how well the agent answers.
Trends over time
Period-over-period movement on every metric, so you know whether a change made things better or worse.
Tool reliability
Track how often each function is called and how often it succeeds — catch a flaky integration before customers do.
Escalation reasons
Understand exactly why conversations reach a human, then close the gaps that drive avoidable handoffs.
Simulations
An agent-eval test suite runs scripted customers and grades the agent — so you ship changes without regressions.
Releases audit trail
A change log of who changed what and when, so every shift in behavior is traceable back to a decision.
Insights console
Know what's working — and what isn't
The Insights overview brings KPIs, period-over-period trends, top customer questions, and tool-usage reliability into one dashboard. Spot a dip in resolution, a spike in a single intent, or a tool that's failing, and act on it the same day.
- Resolution, deflection, and volume at a glance
- Period-over-period trends on every KPI
- Top questions and intents customers bring
- Tool-call reliability and escalation reasons
Simulations
Ship changes without regressions
Simulations is an agent-eval test suite. Scripted customers run against your agent and an LLM judge grades each transcript, with tag and tool-call checks confirming the agent did the right things. Run the suite before every change and catch regressions before customers do.
- Scripted users exercise real conversation paths
- An LLM judge grades each run against your criteria
- Tag and tool-call assertions verify correct behavior
- Tools mocked by default for fast, safe, repeatable runs
Releases
Every change, on the record
The Releases tab is an audit trail of how the agent has evolved — who changed which knowledge, tool, journey, or setting, and when. When behavior shifts, you can trace it straight back to the decision that caused it.
- Who changed what, and exactly when
- Covers knowledge, tools, journeys, and settings
- Tie a behavior change to the edit behind it
- Captured automatically — nothing to remember to log
Frequently asked questions
What does the Insights console show?+
KPIs on volume, resolution, and deflection with period-over-period trends, plus top customer questions, tool-usage reliability, and the reasons conversations escalate — all in one dashboard.
What are simulations?+
An agent-eval test suite. Scripted customers run against your agent and an LLM judge grades each transcript, with tag and tool-call checks confirming correct behavior, so you catch regressions before shipping.
Do simulations call my real systems?+
Tools are mocked by default for fast, safe, repeatable runs. You opt specific simulations into live behavior only where it's appropriate.
How do I know what changed and why?+
The Releases tab is an audit trail recording who changed which knowledge, tool, journey, or setting and when — so any shift in behavior is traceable to a decision.