Noveum

Noveum is a production evaluation platform for chat, voice, and workflow agents. It is best suited to teams that need to diagnose agent behavior and validate fixes, not just collect traces. Its strongest case is the connected path from tracing to reviewable fixes; its practical limits are plan quotas and deployment-specific controls.

Overview

Noveum combines NovaTrace for recording agent activity, NovaEval for scoring it, and NovaPilot for testing proposed corrections. That joined-up workflow can help teams move from a failing production call to a measured, human-reviewed fix without treating tracing and evaluation as separate jobs.

Its intended users are teams running production chatbots, voice agents, or multi-agent workflows. Open-source tracing SDKs support Python and TypeScript, with native integrations for LangChain, LangGraph, LiveKit, Pipecat, and CrewAI, plus compatibility with OpenAI, Anthropic, and OpenTelemetry.

Key features

Tracing and evaluation

NovaTrace captures LLM calls, tool invocations, retrieval, and agent steps, with token use, cost, and latency on spans. This is useful when teams need to pinpoint where an agent flow is spending time or money, rather than seeing only its final response. NovaEval scores traces with calibrated scorers across 15+ categories, including voice and audio; the breadth matters for teams evaluating more than text-only exchanges.

Fix validation

NovaPilot backtests candidate fixes against failing calls, re-simulates them end to end, and delivers validated changes as pull requests for human review. That gives teams a path from identifying a regression to assessing a proposed correction, while keeping the final change reviewable rather than automatic. NovaGuard runtime guardrails are marked beta or coming soon, so teams that need mature runtime protection should look elsewhere for that capability.

Deployment and security

Deployment choices include managed cloud, VPC, BYO ClickHouse, and on-premise, with Kubernetes and Helm options also described. That range can suit organizations with different infrastructure requirements. GDPR and SOC 2 Type II are in progress; enterprise deployments include encryption in transit and at rest and fine-grained access controls. SSO/SAML, RBAC, and audit logging are listed, with pricing specifying SSO/SAML for Scale and above.

Pricing

The Free plan costs nothing and includes 2.5K credits/mo, 1M spans/mo, 2 GB storage, 3 members, and 30-day retention. It is a useful starting point for a small team, but the tight monthly credits and span allowance limit sustained production volume.

Pro costs $69 a month, billed $69/mo, with 10K credits/mo, 20M spans/mo, 20 GB storage, 10 members, and 90-day retention. It includes a 7-day trial. Growth costs $99 a month, billed $99/mo, and raises those limits to 25K credits/mo, 50M spans/mo, 50 GB, 15 members, and 180-day retention; it also has a 7-day trial and priority support. Pro suits smaller teams needing a paid production tier, while Growth is a clearer fit when higher throughput, more seats, and longer retention matter.

Max costs $199 a month, billed $199/mo, and provides 50K credits/mo, 100M spans/mo, 100 GB, 25 members, and 1-year retention, with a 7-day trial. Scale costs $599 a month, billed $599/mo, and includes 200K credits/mo, 500M spans/mo, 500 GB, unlimited members, 2-year retention, and SSO/SAML. Enterprise has custom pricing, with unlimited credits, spans, storage, and members, plus custom retention and dedicated customer success. These higher tiers are aimed at larger or longer-retention workloads; they are hard to justify if the free or Pro quotas already cover the team.

One-time add-ons provide credits without a monthly subscription: Boost is $35, paid once for 5,000 credits, Accelerate is $150, paid once for 25,000, and Scale (credit pack) is $500, paid once for 100,000. Add-on credits never expire, unlike monthly plan credits, which reset each billing cycle and do not roll over. Community support is listed, with priority support for Growth and above. The free tier is the only free plan; paid plans begin at $69/mo.

Platforms

Noveum supports API, Linux, self-hosted, and web deployments. Its mix of managed and self-hosted options is useful for teams balancing operational convenience with infrastructure preferences.

Who it's for

Noveum is a strong fit for teams operating production agents that need trace-level diagnosis, structured evaluation across text and voice, and a repeatable way to validate corrections. It is less suitable for solo or low-volume users who need more than the free limits, or for organizations that require completed GDPR or SOC 2 Type II status or ready runtime guardrails.

Pros and cons

• Pros: Tracing, evaluation, and fix validation form a connected workflow, with validated changes handed off as pull requests for human review.
• Pros: Voice and audio scoring, plus integrations spanning agent frameworks and voice stacks, make it relevant beyond text-only agents.
• Pros: Managed cloud, VPC, BYO ClickHouse, and on-premise options give teams meaningful deployment choice.
• Cons: Free use is capped at 2.5K credits/mo, 1M spans/mo, 3 members, and 30-day retention, so it may not stretch far for production teams.
• Cons: Monthly credits expire at cycle end, requiring teams with uneven usage to plan around quotas or buy non-expiring add-ons.
• Cons: NovaGuard remains beta or coming soon, and GDPR and SOC 2 Type II are in progress.

Alternatives

Browse the AI Agent Evaluation Tools list to compare options. Choose Future AGI AI Evaluation SDK if Apache 2.0 licensing and self-hosting with Docker, Python, or Node SDKs are priorities. Promptfoo is another freemium option with a free Community plan that includes 10k red-team probes/month and local or self-hosted use.

Consider W&B Weave for its free plan with 1 GB/mo of data ingestion, 5 GB/mo storage, and evaluations, tracing, and scorers. DeepEval is a free, Apache 2.0 open-source framework for local and CI/CD evaluation. Choose Opik if you want an open-source core observability and evaluation feature set that can be run locally.

Strands Evals is a free alternative. AgentClash offers a free tier with one workspace, 25 eval runs per month, and 7-day replay retention. Arklex is another free alternative.

Verdict

Choose Noveum if your team runs production chat, voice, or workflow agents and wants one platform to trace failures, score behavior, and validate fixes for human review. Its breadth and deployment choices are persuasive, but quota-sensitive teams should account for expiring monthly credits, and organizations requiring established compliance status or runtime guardrails should look elsewhere.