Cekura

Cekura is an automated quality assurance and observability platform for conversational AI. It enables engineering and QA teams to test, monitor, and improve voice and chat AI agents across the entire deployment lifecycle. The platform is used by companies building on voice infrastructure providers such as Vapi, Retell AI, Bland AI, ElevenLabs, and LiveKit.

How It Works

• Pre-production testing: Teams create agents on Cekura and run simulated conversations across diverse personas, accents, and edge-case scenarios. The platform generates custom evaluation metrics and scores agent performance across dimensions including instruction following, empathy, responsiveness, hallucination detection, and tool call correctness.
• Production observability: Deployed agents are monitored in real time with voice-specific quality signals including gibberish detection, interruption tracking, latency, sentiment, and pitch. Custom dashboards let teams visualize duration trends, sentiment shifts, drop-off rates, and success rates.
• Continuous improvement: Real production conversations can be replayed as regression tests. Teams tune LLM evaluation prompts against recorded calls, diagnose failing evaluations, apply prompt fixes, and catch regressions before updates reach users.

Key Capabilities

• Scenario generation: A library of thousands of pre-built test scenarios covering common conversational flows such as appointment booking, cancellations, and customer support, with support for custom scenario creation.
• LLM-based evaluation: Evaluators score conversations across configurable quality dimensions using predefined and custom metrics. The platform provides prompt recommendations to improve metric scores.
• Real-time alerting: Teams configure thresholds for errors, failures, and performance drops and receive notifications through Slack, email, or webhooks.
• MCP server integration: Cekura provides a Model Context Protocol server that lets AI coding agents such as Claude Code and Cursor trigger test runs, schedule recurring evaluations, and review pass or fail results from their editor.
• Integrations: Direct connections with Retell AI, Vapi, Bland AI, ElevenLabs, LiveKit, Pipecat, Cisco, Five9, and Synthflow, plus CI/CD pipeline integration for automated quality gates.

Use Cases

• Pre-deployment QA: Engineering teams validate new voice agent prompts, workflows, and tool integrations before releasing to production.
• Production monitoring: Operations teams track live call quality metrics and receive automated alerts when latency or error rates exceed thresholds.
• Regression testing: QA teams replay known trouble conversations to verify that agent updates do not reintroduce resolved issues.
• Agent self-improvement: Developers build automated loops where failing evaluations trigger prompt fixes that are tested and deployed, driving continuous improvement toward target pass rates.

Pricing

Cekura offers a free trial. Paid plans are custom and require contacting sales.

Privacy and Security

Cekura is SOC 2 and HIPAA compliant and adheres to GDPR requirements. The company is backed by Y Combinator and has raised $2.4 million in funding.