The Serenity* Star Agent Platform
Design, deploy, and orchestrate AI agents at enterprise scale.
Why most AI projects fail
AI adoption is often slowed down by complex integrations, reliance on specialized AI expertise, and scalability issues. As enterprises scale AI, they encounter new obstacles in governance, compliance, and performance visibility.
- Experimentation to Production Gap
AI demos and pilots are easy to build, but scaling to production is hard. Unlike controlled experiments, production AI must navigate complex rules, exceptions, and legacy processes. This gap often leads to stalled initiatives that fail to deliver real business value.
- Enterprise Context Integration
AI agents need access to business context to be effective, but enterprise knowledge is scattered across disconnected systems. Integrating these sources requires significant technical expertise, which many organizations lack. Without these connections, AI agents operate with limited context.
- AI Governance & Compliance
Deploying AI at enterprise scale demands robust governance frameworks that ensure appropriate use, maintain audit trails, and enforce security standards. Companies must implement mechanisms to prevent data leakage, manage access controls, and comply with industry-specific regulations like HIPAA, GDPR, or financial services requirements.
- Operational Scalability
Moving from AI pilots to production requires AI systems that can handle enterprise volumes while maintaining performance and reliability. Companies struggle to design agent architectures that can scale horizontally, manage peak loads, handle complex workflows across multiple agents, and integrate with existing operational monitoring.
Why Serenity* Star for Production-grade AI
The Kore.ai Agent Platform is an enterprise-grade foundation for building AI agents at scale, transforming service, productivity, and business processes. The platform provides a unified framework with robust security, enabling rapid AI implementation while maintaining enterprise control.
Simplify everyday business tasks.
Because you’d probably be a little confused if we suggested you complicate them instead.
Dynamically break down complex user queries into structured subtasks with AI agents, selecting optimal search parameters and filters at each step to retrieve the most relevant, context-aware information.
Store and recall context from ongoing conversations with short-term memory and retain user-specific details to rephrase user questions with long-term memory.
Comprehensive system for connecting agents with enterprise applications, featuring 100+ pre-built deep integrations and visual tools to develop custom integrations without extensive coding.
Builder & Design
Visual Agent Designer
Design assistants, copilots, and activity agents with an intuitive visual experience. Configure model, behaviour, skills, security, parameters, and knowledge—test changes instantly and publish when ready.
Pick LLMs and tune parameters (max tokens, temperature, top‑p). Define system prompts, initial messages and conversation starters.
Attach skills (function calls), add input parameters using {{tokens}}, and restrict operations via metacontrols and allowed domains.
Upload knowledge, iterate safely in drafts, then publish versions. Preview changes before rolling out to production.
Knowledge & RAG
Bring your own websites and files. Serenity processes, segments, embeds and indexes content for high‑quality retrieval‑augmented generation.
Add multiple URLs or documents. Content is extracted, chunked with overlaps and embedded automatically.
Inspect and edit sections, adjust pre/post overlaps and delimit long docs with custom separators.
Schedule reprocessing for websites, track status, and tie changes to draft agent versions for safe publishing.
Skills (Function Calling)
Extend agents with business logic by exposing APIs and utilities as callable functions. Use schemas so models know how to call and interpret them.
Connect CRMs, ERPs or internal services without rewriting code. Reuse skills across multiple agents.
Support retrieval functions that summarise data and task functions that perform side‑effects like creating tickets.
Define JSON schemas for inputs and outputs so the model calls tools correctly and responses are easy to validate.
Workflow Orchestration
Coordinate specialised agents and background activities to deliver outcomes. Support branching, delegation and event‑driven flows.
Compose assistants, activity agents and copilots into virtual teams to solve complex problems end‑to‑end.
Route by intent, confidence, cost or latency. Apply governance policies at every step.
Run background jobs for enrichment and long‑running actions while keeping the conversation responsive.
Governance & Safety
Policy Metacontrols
Enforce responsible use with domain allow‑lists, guardrails and moderation. Keep critical workflows safe and compliant.
Limit what agents can fetch or execute using allowed domains and strict policy rules.
Apply enterprise security standards and moderation consistently across agents and environments.
Introduce human‑in‑the‑loop for sensitive actions and regulated domains when needed.
Audit & Versioning
Track configuration changes, test safely in drafts and publish with confidence. Version explicitly to avoid surprises.
Edit safely in drafts; publish new versions when validated by your team.
Maintain audit trails of knowledge edits, parameter updates and policy changes.
Knowledge sources are tied to versions, enabling reproducible results over time.
Agent Quality Studio
Plan, execute and evaluate test suites to validate agent behaviour before release and over time.
Define cases per intent or scenario with expected outputs and criteria.
Run plans and review reports with pass/fail results and quality metrics.
Drill into failures, refine prompts/knowledge/skills and compare historical runs.
Privacy & Masking
Protect sensitive data and uphold ethical AI principles with proactive controls and transparency.
Avoid sending PII or confidential data; enable masking and redaction where appropriate.
Prevent the generation of harmful content and provide clear disclosures to users.
Support GDPR and other regulations with auditability and policy enforcement.
Runtime
Multi‑Model Routing
Choose the right model for the job—balance speed, cost and quality with policy‑driven routing across providers.
Connect to providers like Azure OpenAI, Anthropic, Groq, Google, Mistral and more—no lock‑in.
Route by intent, size or latency budget; set fallbacks and guard against rate limits.
Apply consistent controls and observability regardless of the underlying model.
Conversation Memory
Maintain context across turns and reuse chat IDs to preserve state throughout a session.
Preserve prior messages to keep conversations coherent and personalised.
Tune message windows and summarisation strategies for optimal quality and cost.
Reuse chat IDs for the lifetime of the conversation as recommended in best practices.
Streaming & Tools
Deliver fast, fluent user experiences with streaming responses and robust tool (function) calling.
Render tokens as they arrive for a responsive chat UX similar to Chat Completion APIs.
Expose business functions with validated inputs/outputs and secure execution paths.
Collect structured data via forms and let users upload files that agents can use.
Background Jobs
Run activity agents asynchronously for long‑running tasks, enrichment and automations outside the main chat loop.
Trigger side‑effecting actions without blocking the conversation or UI.
Report status back into the chat once jobs complete or when insights are ready.
Scale reliably for enterprise workloads and peak periods.
Observability
Analytics & Dashboards
Track usage, latency, token consumption and model costs to optimise quality and spend.
Monitor throughput, response time and error rates across agents and models.
Use token usage and billing data to control budget and forecast spend.
Spot regressions and improvements over time with clear visualisations.
Conversation Sampling
Analyse a percentage of conversations automatically and extract custom fields for decision‑making.
Set analysis sample rates and define additional instructions for extraction.
Create fields like sentiment or outcome and see them populated by the model.
Access a grid of analysed conversations and drill into details as needed.
Trace Explorer
Inspect end‑to‑end executions to understand how prompts, tools and knowledge produced an answer.
Follow messages, tool calls and knowledge lookups step‑by‑step.
Identify bottlenecks and failure points to improve reliability.
Maintain auditability while respecting data protection policies.
Cost Controls
Keep operations within budget using subscriptions, credits, auto‑recharge and real‑time usage monitoring.
Top up manually or set thresholds that trigger automatic purchases via Stripe.
Choose a plan, manage add‑ons and handle renewals with grace periods on failures.
Use dashboards and APIs to track token spend per agent and environment.
Acceleration
Pre‑built Copilots
Accelerate delivery with ready‑to‑use assistants and templates that you can customise for your domain.
Embed web chat or React components in minutes and configure with your agent code.
Starter designs come with sensible defaults for prompts, memory and guardrails.
Adjust tone, add skills and connect knowledge to fit your organisation.
SDK & Components
Integrate with lightweight SDKs or drop‑in UI components. Use REST APIs directly when you need full control.
Create conversations, execute messages and upload files with a simple client.
Use the Serenity Chat component with theming, streaming and forms support.
Call REST endpoints from any server and keep API keys secure.
Evaluation Templates
Validate quality with reusable test sets and scoring strategies powered by Agent Quality Studio.
Share baseline test plans across teams and compare runs over time.
Track accuracy, completeness and other custom metrics for your use cases.
Catch regressions before release by integrating evaluation into your workflow.
Deployment Flexibility
Run Serenity as SaaS, in your VPC or on‑premises with consistent security and governance.
Choose the model that matches your compliance and operational needs.
Bring your own models and infrastructure—avoid lock‑in and keep options open.
Apply the same policies, guardrails and observability across environments.