Overview

Arize AX combines agent observability with tools for investigating failures, evaluating behavior, and testing changes. It is best suited to teams operating agents across several frameworks, especially those that need to follow issues from production traces through to fixes and experiments. Its breadth is compelling, but teams should match its trace, ingestion, and retention quotas to their telemetry needs before choosing a plan.

Key features

Signal surfaces issues, investigates root causes, and produces fixes for review. Managed Agents can examine failing traces, address repository bugs, automate recurring checks, or be built from scratch. That makes AX a fit for teams seeking both diagnosis and follow-up work in one platform; the monthly Signal caps on Free and Pro may constrain teams that investigate many issues.

Swarm Observability tracks the real-time status and performance of an agent swarm. Agent-as-a-Judge builds scalable evaluations from agent telemetry to help expose failures teams did not anticipate, while end-to-end experiments let teams validate agent changes. This linked monitoring, evaluation, and experimentation workflow is more useful for continuous improvement than tracing alone.

Tracing integrations cover OpenAI, Anthropic, Google GenAI, Amazon Bedrock, LangChain, LangGraph, LlamaIndex, CrewAI, and Vercel AI SDK. OpenInference conventions and OpenTelemetry support instrumentation across these frameworks and providers. AX also supports LLM, agent, and retrieval tracing, evaluations, prompt management, and token cost tracking, making it relevant to teams that need to inspect both agent behavior and model usage.

Self-hosted Enterprise keeps observability data, authentication, APIs, and the UI within a customer’s infrastructure, with air-gapped deployment available. That is a meaningful option for organizations with strict infrastructure requirements, though it requires the Enterprise plan rather than the SaaS-only Free or Pro tiers. Arize lists SOC 2, PCI DSS, ISO/IEC 27001, GDPR, and HIPAA compliance or certification in its Trust Center.

Pricing

Plan	Price	Limits and deployment	Support
AX Free	0.00 USD per free	10 Signal issues/month; 25k trace spans/month; 1 GB ingestion/month; 15-day retention; SaaS; unlimited users and evaluations	Community troubleshooting
AX Pro	50.00 USD per month	25 Signal issues/month; 50k trace spans/month; 10 GB ingestion/month; 30-day retention; SaaS; unlimited users and evaluations	Email
AX Enterprise	Custom pricing	Unlimited Signal; custom trace spans, ingestion, and retention; SaaS or self-hosted; unlimited users and evaluations	Dedicated

Free is a practical starting point for smaller workloads: it has no seat or evaluation cap, but its 25k monthly spans, 1 GB ingestion, 15-day retention, and 10 Signal issues limit how much production activity teams can retain and investigate. Pro costs 50.00 USD per month and doubles the span allowance, raises ingestion to 10 GB, and extends retention to 30 days, while increasing Signal to 25 issues. It keeps unlimited users and evaluations, so the upgrade is principally about telemetry capacity and investigation volume rather than seats. Enterprise is the option for custom volume and retention, unlimited Signal, or self-hosting; its price is custom. Free and Pro are SaaS deployments, while all three plans list US, EU, or CA data regions.

Platforms

AX is available through API and web access, with self-hosted deployment on Enterprise. The SaaS plans suit teams that prefer a managed deployment; organizations requiring observability data, authentication, APIs, and UI inside their own infrastructure can consider Enterprise, including an air-gapped deployment.

Who it's for

AX fits teams building production AI agents that need tracing across established model providers and agent frameworks, plus evaluations and experiments to improve behavior over time. Its swarm monitoring and issue-investigation workflow particularly suit teams coordinating multiple agents. Teams with modest telemetry can start on Free; teams needing more retention and ingestion can move to Pro. It is less suitable when a team needs self-hosting but cannot pursue custom-priced Enterprise, or when its fixed lower-tier quotas are too restrictive.

Pros and cons

Pros: Tracing integrations span major model providers and agent frameworks, with OpenInference and OpenTelemetry conventions for instrumentation.
Pros: Investigations, telemetry-driven evaluations, swarm monitoring, and end-to-end experiments connect several stages of agent improvement.
Pros: Free and Pro include unlimited users and evaluations, avoiding seat-based restrictions as a team grows.
Pros: Enterprise supports self-hosting, air-gapped deployment, and custom telemetry limits.
Cons: Free caps retention at 15 days, ingestion at 1 GB/month, spans at 25k/month, and Signal at 10 issues/month.
Cons: Pro raises those quotas but remains capped at 50k spans, 10 GB ingestion, 30-day retention, and 25 Signal issues monthly.
Cons: Self-hosting and custom quotas require Enterprise, which has custom pricing.

Alternatives

For a broader shortlist, compare LLM Observability Tools, AI Agent Observability Tools, AI LLM Observability Tools, Machine Learning Model Monitoring Software, and Model Monitoring Software.

Evidently AI is worth considering for teams that prioritize an Apache 2.0 open-source framework and broad platform availability, including desktop and Linux options.
Helicone is another freemium, open-source option for teams seeking API, self-hosted, and web access.
Laminar may suit a small team that values a free plan with one seat, one project, 1 GB data, and seven-day retention.
Langfuse offers a free Hobby tier with 50k units per month, 30 days of data access, and two users; consider it when those limits fit your team.
Opik is an alternative with an open-source core observability and evaluation feature set that users can run locally.
Respan (formerly Keywords AI) is a freemium alternative whose Free plan includes 100k logs, 1k scores, five datasets, two evaluators, and five prompts.
W&B Weave may fit teams seeking AI application evaluations, tracing, and scorers within a free tier that includes 1 GB/month of data ingestion and 5 GB/month of storage.
Braintrust is an alternative with a free Starter plan including 1 GB processed data, 10,000 scores, 14-day retention, and unlimited users, projects, and datasets.

Verdict

Choose Arize AX if your team needs a connected agent workflow for tracing, issue investigation, evaluations, and experiments, with integrations across major providers and frameworks. The free tier is useful for a bounded workload, and Pro adds capacity without seat limits; look elsewhere if you need self-hosting without custom pricing or expect to exceed the lower tiers’ telemetry and Signal quotas.

	Arize AX	Lunary
Free plan	Yes	No
Free trial	No	No
Paid from	$50/mo	—
Open source	No	No
Platforms	api, self-hosted, Web	Web
Free plan	Yes	No
Paid from	50 /mo	—
Cost tracking	Yes	Yes

Arize AX vs Lunary

Arize AX

Lunary