Arthur

Overview

Arthur discovers AI agents across enterprise environments, then gives teams a shared way to observe their activity and enforce policies. It suits organizations managing agents across multiple systems, especially those with deployment or data-control requirements. Its strongest case is the combination of discovery, monitoring, and governance; the Free plan’s short retention and project cap make it better for a bounded evaluation than broad ongoing monitoring.

Key features

Discovery sensors watch OpenTelemetry streams, MCP servers, network traffic, cloud APIs, and employee endpoints. That broad set of entry points is useful when agents are spread across an organization rather than built within one framework, though the value depends on the environments a team needs to cover.

Arthur traces agent steps, tool calls, retrieved context, costs, latency, and behavior over time. Teams can use those signals to investigate how an agent reached an outcome and track operational changes. Drift monitoring, model performance metrics, and bias monitoring extend the picture beyond individual traces; webhook and Slack alert channels can carry notifications into existing workflows.

Governance combines machine-checkable policies with real-time guardrail verdicts for prompt injection, PII exposure, restricted models, and off-policy tool use. This is a more direct fit for teams that need controls during agent activity, not just a record of it. Documented integrations span OpenAI, Anthropic, LangChain, LiteLLM, CrewAI, LlamaIndex, OpenAI Agents, Google ADK, AWS Bedrock, and Mastra. Data connectors include BigQuery, Google Cloud Storage, Amazon S3, and Arthur Shield instances.

Arthur’s open-source Evals Engine includes built-in PII, sensitive-data, custom-LLM, and regex rules, with self-serve deployment. In hybrid deployments, its evals engine runs next to customer workloads so sensitive data can remain in the environment. Hosted SaaS, on-premises, and hybrid options give organizations room to choose their operating model, while RBAC, SSO, and deployed engineering support come with every deployment option.

Pricing

Arthur is freemium, with paid plans from $60/mo. The Free plan costs nothing (billed $0/mo) and covers up to 4 use cases, 1 organization, 1 workspace, 2 projects, 7-day data retention, 5k jobs, 300k spans, 12k inferences, and 3k evals. Unlimited seats make it accessible to a team, but the two-project cap and one-week retention constrain broader rollouts and historical investigation.

Premium costs $60 a month (billed $60/mo). It raises the limits to 100 use cases, 10 projects, 30-day retention, 20k jobs, 1.2M spans, 100k inferences, and 75k evals, while retaining 1 organization and 1 workspace. That is a substantial capacity increase for a team whose monitoring fits within one organizational space, but retention remains limited to 30 days.

Enterprise has custom pricing. It removes fixed limits on organizations, workspaces, projects, and data retention, and uses custom jobs, spans, inferences, and evals. Dedicated and managed VPC options, a dedicated customer success manager, advanced monitoring, SSO, SLAs, and BAA make it the relevant tier for organizations with deployment, support, or compliance requirements that exceed Premium’s bounds.

Platforms

Arthur is available through API, self-hosted, and web options. Hosted SaaS, on-premises, and hybrid deployments address different infrastructure needs; the hybrid evals engine is particularly relevant when sensitive data must stay alongside customer workloads.

Who it's for

Arthur is best suited to enterprises that need to find agents across varied environments, inspect their behavior, and apply runtime policy controls from a common platform. Its deployment choices and integrations suit organizations with established cloud, framework, and data systems. A small team monitoring one or two projects may find the Free plan useful, but teams that need longer retention or larger quotas will need Premium or custom-priced Enterprise.

Pros and cons

• Pros: Discovery across streams, servers, network traffic, cloud APIs, and endpoints can help with agent visibility beyond a single development framework.
• Pros: Tracing, drift and performance metrics, bias monitoring, and real-time policy verdicts bring operational oversight and guardrails together.
• Pros: Hosted, on-premises, and hybrid deployment options accommodate different infrastructure and data-handling needs.
• Cons: Free retains data for only 7 days and is capped at 2 projects, limiting its usefulness for sustained or multi-project monitoring.
• Cons: Premium still limits customers to 1 organization and 1 workspace, while Enterprise requires custom pricing.

Alternatives

For readers comparing model monitoring tools, see Machine Learning Model Monitoring Software and Model Monitoring Software.

Deepchecks is worth considering for a free starting point capped at 3 seats and 1 AI application, with 3 months of retention and unlimited prompt-based metrics. SUPERWISE may suit an individual trying a paid service: its Solo plan is $10 a month after 30 days free and covers one Sentinel deployment for development use. NannyML offers a self-managed free Open Source plan, while its Starter plan is $399 a month for 2 models and 10 M predictions.

Arize AX may appeal to teams seeking a free SaaS tier with unlimited users and evals, but its AX Free plan is capped at 25k trace spans per month, 1 GB ingestion per month, 10 issues per month, and 15-day retention. Opik makes its core observability and evaluation features available as open-source code for local use.

Evidently AI is an option for teams that want a framework fully open-source under the Apache 2.0 license. Galileo offers a Pro plan at $100 a month billed yearly, with 50,000 traces per month, standard RBAC, advanced analytics and insights, and dedicated Slack support.

Verdict

Choose Arthur if you need enterprise-wide agent discovery alongside trace-level observability, runtime governance, and a choice of hosted, on-premises, or hybrid deployment. Its breadth is the main reason to choose it; its short Free retention, Premium’s single-organization limit, and custom-priced Enterprise tier are reasons to look elsewhere if your needs are smaller, more narrowly scoped, or price-sensitive.