Product Updates | March Week 1, 2026

⚙️ Auto-Sync Retell Agent

Say goodbye to the copy-paste grind. You can now automatically sync all Retell agent settings directly. Any updates to your prompts or configurations reflect instantly on Cekura, keeping your workflow unified and error-free.

---

🌁 On-Prem Observability Setup

For teams requiring maximum data sovereignty and low-latency monitoring, we now offer On-Premise setup for our observability suite. Keep your sensitive interaction data within your own infrastructure while maintaining full visibility into agent performance.

Note: Please reach out to us via any support channel or reply to this email if you are interested in this setup.

---

📋 Test Profiles as Dynamic Variables

We will now automatically pass test profile data as dynamic variables across multiple triggers, including:

• Platforms: Retell, VAPI, and ElevenLabs (WebRTC/Auto-outbound).
• Protocols: WebSocket Chat headers and SIP headers.

---

☄️ Other Improvements

• BigQuery Integration: Sync your knowledge base directly to Cekura.
• CSAT Sentiment Capture: The CSAT metric now captures the sentiment of the user as well!
• View-Only Members: Invite stakeholders to monitor progress without risking configuration changes or paying for extra seats.
• Chat in Action Items: Speak with your evaluation data directly within the action item interface.

⚙️ AUTO-SYNC MOCK TOOLS

You can now automatically mock your agent’s tools during testing with Auto-Sync Mock Tools. The platform fetches all tools configured for your agent on Retell / VAPI / Elevenlabs, temporarily replaces them with mock endpoints for safe testing, and restores the original configuration once testing is complete.

This removes the need to manually mock tools or configure values in your CRM and makes testing faster and more reliable. Simply go to Agent Settings → Mock Tools, use Auto Fetch to generate mocks, and revert back to the original setup anytime with a single click.

🌁 VIEWS

You can now create and save custom Views to quickly access calls and results for the agents you care about. By default each agent is available as a view. Views extend the previous per-agent filtering by allowing you to group multiple agents together, making it easier to monitor Results and Calls across teams or environments in one place.

☄️ OTHER IMPROVEMENTS

Silence during a turn - Structured test cases now support remaining silent during a turn. Reach out if you need help configuring it.

Default Enable/Disable for Metrics - You can toggle individual metrics for specific agents. There is also a new setting to enable or disable a metric for any new agent created .

👍🏻 FEEDBACK FROM SLACK

You can now provide feedback directly by using the thumbs-down button on alerts in slack. This ensures you don't have to leave Slack to share your input, and it routes your feedback directly to the Labs.

⚠️ REVIEW REQUIRED STATE

You can now easily identify conversations where the system could not determine a conclusive True or False outcome. This feature flags calls for your manual attention when automated evaluation isn't possible, such as when a call ends prematurely or the testing agent behaves unexpectedly.

☄️ OTHER IMPROVEMENTS

Billing Page - The billing page is completely revamped with clear usage charts and breakdown per metric, easy visibility into the next bill..

Improved Integration with VAPI/Retell/Elevenlabs - You can now enable auto-fetch for VAPI/Retell/Elevenlabs agents for production monitoring. We automatically fetch calls and relevant metadata to provide you great insights.

🪾 IVR TESTING

You can now create dedicated test cases to simulate IVR interactions where testing agents are IVR trees. This feature is fully configurable, allowing you to full control over the IVR menu and also utilize DTMF tones (like press 1) to ensure your system can navigate IVR Menus correctly. Read more in IVR & Voicemail Docs.

🔧 MOCK TOOLS

You can now eliminate production dependencies and unblock your testing workflows with Mock Tools. This feature elevates Cekura from a platform to a fully self-contained environment, allowing you to mimic external tool calls seamlessly. Simply define the tools you use, and we will generate and verify the input and output for those tool calls automatically. Go to Agent Settings to find this option.

🪜 VOICEMAIL TESTING

You can now create dedicated test cases to simulate voicemail interactions. This feature is fully configurable, allowing you to define exactly when the beep plays and utilize DTMF tones (like press #) to ensure your system handles voicemail drops and detection logic correctly. Read more on Voicemail Testing.

🔌 TELNYX PSTN INTEGRATION

We have expanded our "Bring Your Own Telephony" capabilities to include Telnyx. Just as with Twilio, you can now seamlessly link your Telnyx numbers, allowing you to leverage your existing carrier rates and relationships within our system.

🧵 OTHER IMPROVEMENTS

• Dashboards - Updated dashboards now let's you filter and group by based on other fields like metric value, metadata etc. Leverage it to easily build your A/B comparison dashboards

• CI/CD Test Suite - We have added a pre-built suite of test cases for Infrastructure testing like "Saying first message", "Following up on no response" etc. To get an early access reach out to the team!

🎁

The lights are twinkling, the cocoa is hot, and we’re sliding into the final stretch of the year. It's the holidays and we are wrapping up the year with our last update for the year.

🔐 RED TEAMING SCENARIOS

This is the big one. We’re giving you the ultimate "armor" for your agents with our new Red Teaming suite. Designed to stress-test your AI’s security and safety boundaries, this feature goes far beyond simple checks.

Leveraging our proprietary research, we’ve curated a massive library of 10,000+ specialized scenarios to ground every test. These aren't just one-off prompts; they are specifically built for multi-turn conversations, mimicking how real-world bad actors or complex edge cases actually behave. It’s the most robust way to ensure your agent stays on track, secure, and resilient - no matter what is thrown its way.

* ‍​Generate ‍​Evaluators ‍​-> ‍​Scenario ‍​Type ‍​-> ‍​Red ‍​Teaming

📊 CUSTOM OBSERVABILITY DASHBOARDS

Bring your data to life with our brand-new Custom Dashboards. This feature gives you total visibility into your AI conversations, allowing you to move beyond raw logs and see the big picture of your agent's performance.

Whether you need a high-level overview or a deep dive, you can now plot specific metrics and visualize them as lists, bar graphs, and more. We’ve also included powerful "Group By" filters, making it easier than ever to run side-by-side A/B testing comparisons. Simply go to observability hub to give it a try.

🪜 TRANSCRIPTION ACCURACY METRIC

Production calls are failing constantly due to poor transcription? With our new Transcription Accuracy metric you can easily track poor transcription issues, benchmark different STT providers and receive alerts when performance deteriorates significantly.

🔌 SIP INTEGRATION

You can now connect directly to your agent over SIP. This enables seamless, real-time interoperability with your existing telephony systems. We also send run ids and other metadata as SIP headers to help you match runs to your internal logs. Read more on SIP Integration.

📞 BRING YOUR OWN PHONE NUMBER

Need phone numbers from regions outside the US? Want exclusive numbers for your outbound testing workflows? Have internally whitelabeled numbers you rely on? You can now bring your own phone numbers. We’re starting with full support for any Twilio number, and we’re happy to expand - just reach out if you're using a different provider.

* ‍​Settings ‍​-> ‍​Telephony ‍​-> ‍​Import ‍​Numbers

🪜 PROJECT METRICS

Easily create metrics or move existing agent metrics to be Project Metrics. These metrics are applicable across all agents you select, but edited from one source making metric management seamless. They also share 1 lab.

🔔 TREND BASED ALERTS

You can now set up intelligent alerts via Slack or email that notify you when metrics drift from their normal patterns. Instead of setting rigid limits, this feature detects unusual spikes or drops based on recent activity, helping you catch anomalies automatically.

* ‍​Settings ‍​-> ‍​Notifications ‍​-> ‍​Observability ‍​Alerts

🔧 OTHER IMPROVEMENTS

SMS support during call : Test Agents can now receive SMS between call and have full context of the message received. This is very useful to test cases where you are using SMS based 2FA during a call.

Test Profile in Expected Outcome : Added support for using test profile values in metrics, allowing usage of {{test_profile.variable_name}} for more templated evaluations

Integrations : We now support Agentforce chatbot integration and Kore AI voice agents.

🔌 INTEGRATIONS

We have added new integrations to make testing your agents easier for you.

1. Pipecat WebRTC: Just provide your pipecat details and we can automatically create rooms and connect to your agent. All via the UI. Details here.

2. ElevenLabs Websocket: We can now connect to your elevenlabs agents via websockets for voice conversations instead of telephony. Details here.

🌎 REAL-WORLD SIMULATIONS

Found yourself wanting to create test cases from production calls? We have significantly improved our real-world simulation feature.

* ‍​Add ‍​Calls ‍​to ‍​Observability ‍​-> ‍​Click ‍​Actions ‍​-> ‍​Create ‍​a ‍​Simulation ‍​

⛔️ REDACTION FOR OBSERVABILITY

Have sensitive data in your production calls you are worried to share? You can now enable redaction while sending calls to us. Note: We redact sensitive data from transcript and audio both!

Details here.

🔧OTHER IMPROVEMENTS

Automated Outbound Testing : Retell/VAPI/Elevenabs users can now test outbound voice agents easily without having to copy paste phone numbers provided by Cekura. Simply go to

* ‍​Agent ‍​Settings ‍​-> ‍​Voice ‍​Integration ‍​-> ‍​Choose ‍​provider ‍​as ‍​  ‍​ ‍​Retell/VAPI/Elevenlabs ‍​-> ‍​Outbound ‍​Auto ‍​Call ‍​

BLOGS 📖

Complete Chatbot Testing Guide

🎖️ METRIC OPTIMISER

Our metric optimiser is extremely powerful now. Simply downvote on metrics for evaluated conversations, leave a note and click optimise in the lab.

• click downvote -> leave a note -> go to lab -> optimise the metric

⫘ LIVEKIT INTEGRATION

We now support a no-code livekit integration for testing. Simply add details of your livekit account in agent settings and Cekura will create the required rooms in your livekit service and connect to it over webrtc for testing.

• Agent settings -> voice integration -> provider Livekit

📣 NEW AUDIO METRICS

We now have 3 new pre-defined metrics that evaluate on audio.

1. Pronunciation Check - Simply add a word and it's pronunciation via phoneme and we will flag calls where pronunciation breaks. Specifically useful in monitoring how your AI Agent is pronuncing brands.

2. Silence Detection - Detect silences in a conversation which are greater than N (default 10) seconds.

3. Voice Tone - This metric checks for clarity in your AI Agent's voice and any discrepancy in the tone.

🔧 OTHER IMPROVEMENTS

Tool Call Testing : The new improved tool call testing is extremely easy to setup and dramatically expands the coverage of your simulations. If you are interested in tool call testing, refer to this guide or reach out to us.

Custom personalities : Enjoy a much more robust experience to create custom personalities. You can control interruptions, background noises, speed of speaking and much more. Note - we have also improved our pre-defined set of personalities in all language.

Real World Simulation : This is now significantly improved in text. Over the coming week we'll improve it in voice simulations as well.

BLOGS 📖

Test New LLM Model Versions with Real Production Calls

🎖️ SMS BASED TESTING

You can now test your chat-based agents via SMS to ensure they perform reliably in a text-messaging environment.

• Agent Settings -> Chatbot integration -> Choose provider as SMS

📜 RUNS OVERVIEW - DASHBOARDS

Create your own custom dashboards in Runs Overview. This will help you track the performance of your key metrics over time. Specifically handy if you run scheduled tests.

• Runs Overview -> Select Filters -> Save

🔧 OTHER IMPROVEMENTS

Reports from multiple Results : You can now club multiple results to create 1 report.

Downvote Metrics : If any metric value is not accurate, you can downvote it. The team is constantly improving them and this will be help us fine-tune the metrics for your use case.

Stability Fixes : We have optimised load times by >50% to ensure you have a snappy experience.

CASE STUDY

🔔 How Quo is shipping fast via Cekura

BLOGS 📖

Why single turn testing falls short in evaluating conversational AI

🔩 IMPROVED RESULTS PAGE

We've completely overhauled the results page to give you a clearer, more actionable overview of your simulation. You can now quickly identify key insights and, with the new Action Items feature, instantly turn those insights into tasks.

🎭 SCENARIO TYPES: BIAS, TOXIC...

You can now select from a variety of scenario types - including Bias, Toxic, Hallucination, and Sad - to generate more specific and relevant scenario features

Evaluators -> Generate Evaluators -> Scenario Type field

🔧 OTHER IMPROVEMENTS

Alerts for Specific Metric: You can now enable slack/email alerts on each individual metrics.

Slack Notifications for Results : Receive slack notifications once simulation runs are complete.

NEW BLOGS 📖

🔔 12 Supporting Metric to Evaluate your conversational AI Agents

🎖️ AUTO METRIC OPTIMISER

We now automatically optimise metrics. Feedback can be direct or based on feedback. This reduces manual effort while continuously improves accuracy.

• Select Create Metric -> Type Description -> Click Improve

• Labs -> Optimiser -> Metric to Optimise -> Add Note & Improve

🎭 FEEDBACK FOR INSTRUCTION FOLLOW

You can now leave feedback on calls where the agent didn’t follow instructions as expected. This helps us improve evaluation accuracy and capture real-world issues faster.

Failed Calls -> click 👎🏻 -> leave feedback

🔧 OTHER IMPROVEMENTS

Bland Integration : We now support a native integration with Bland for voice agent testing.

Extended Knowledge Base Support : We now support JSON, csv etc.

Beta Metric : Appropriate Call Termination by AI Agent

Evaluator Optimiser : Similar to metric optimiser, when creating scenarios, you can use AI Assist so you don't have to type every detail.

NEW BLOGS 📖

🔔 Choosing the right LLM for your Conversational AI Agent