# Noveum.ai

> Noveum is the AI-native reliability platform for production AI agents and LLM apps. It traces every run, evaluates with 106 calibrated scorers (including 17 dedicated voice scorers), simulates whole scenarios with tool virtualization, guards live traffic, and ships validated fixes as pull requests. Python & TypeScript SDKs; SOC 2 Type II (in progress), GDPR; cloud, VPC, or fully on-premise with BYO ClickHouse.

## Products
- [Tracing — NovaTrace](https://noveum.ai/novatrace): Open-source SDK that captures every LLM call, tool, RAG step, and agent hop as a structured trace. Integrates with LangChain, LangGraph, LiveKit, Pipecat, and CrewAI in a few lines.
- [Chatbot Evaluation — NovaEval](https://noveum.ai/novaeval): Scores chat agents with 106 calibrated LLM-as-judge scorers across 18 categories; every verdict comes with its reasoning.
- [Voice AI Agent Evaluation & Testing — NovaSynth](https://noveum.ai/novasynth): Forward-simulates real voice (SIP) and chat scenarios with personas, accents, interruptions, and tool virtualization to validate end-to-end outcomes.
- [Autonomous Fixing — NovaPilot](https://noveum.ai/novapilot): Reads failing evals, isolates root causes, and ships fixes as pull requests — each backtested per-call and validated end-to-end in simulation.
- [Guardrails — NovaGuard (Beta)](https://noveum.ai/novaguard): Real-time policy enforcement that blocks bad outputs before users see them.

## Solutions
- [AI Agent Monitoring](https://noveum.ai/solutions/ai-agent-monitoring)
- [LLM Observability](https://noveum.ai/solutions/llm-observability)
- [Agent Evaluation](https://noveum.ai/solutions/agent-evaluation)
- [Debugging & Tracing](https://noveum.ai/solutions/debugging)
- [Scorer Library — 106 calibrated evaluators](https://noveum.ai/solutions/scorers)

## Comparisons
- [Noveum vs Langfuse](https://noveum.ai/comparison/noveum-vs-langfuse)
- [Noveum vs Arize & Braintrust](https://noveum.ai/comparison/noveum-vs-arize-braintrust)
- [Noveum vs Arize Phoenix](https://noveum.ai/comparison/noveum-vs-arize)
- [Platform comparison hub](https://noveum.ai/comparison)

## Developers
- [MCP server for AI agents](https://noveum.ai/mcp): Connect Claude, Claude Code, Cursor, VS Code, ChatGPT, Windsurf, Cline, Replit, and any MCP client to Noveum through one remote endpoint (noveum.ai/api/mcp). ~60 tools, 16 read-first resources, and 20 guided workflow prompts, over OAuth 2.1 sign-in or a Bearer API key. Drive tracing, evaluation, cost optimization, and NovaPilot fixes in natural language.
- [MCP server reference](https://noveum.ai/docs/platform/mcp-server-reference): Full tool/resource/prompt catalog, OAuth connector mode, scopes, and per-client install snippets.
- [Noveum Agent Skill](https://github.com/Noveum/noveum-skill): Open-source Agent Skill (SKILL.md) that lets Claude Code, Cursor, or any coding agent set up Noveum end to end in your own environment — SDK integration, trace-completeness verification, datasets, evals, NovaPilot diagnosis, and fix application, with acceptance checks per step. Docs: https://noveum.ai/docs/platform/agent-skill

## Company & resources
- [About Noveum](https://noveum.ai/about): Why we built an AI-native reliability layer.
- [Enterprise](https://noveum.ai/enterprise): On-prem & self-hosted, BYO ClickHouse, SOC 2 Type II (in progress), GDPR.
- [Pricing](https://noveum.ai/pricing)
- [Documentation](https://noveum.ai/docs)
- [Blog](https://noveum.ai/blog)
- [Changelog](https://noveum.ai/changelog)
- [Case study: MyOperator](https://noveum.ai/case-studies/myoperator): Voice-agent call success from 84% to 95%+ with 200× faster iteration.
- [GitHub — noveum-trace SDK](https://github.com/Noveum/noveum-trace)