Arthur
AI observability platform monitoring model performance, bias, drift and behavior for ML and generative AI/agentic applications in production, built on an open-source evals engine.
What it does
- Evaluation and observability
Features 20
- Agent inventoryShows every agent in one inventory, including shadow agents.
- Step-level tracesTraces reasoning steps, tool calls, retrieval and handoffs.
- Policy enforcementEnforces global and use-case policies across agents.
- Runtime securityAgent Behavioral Analytics monitors AI applications and alerts on anomalous behavior.
- Cost controlTracks spend by team and user and reroutes requests that do not need the expensive model.
- GuardrailsGuardrails are the Arthur GenAI Engine's real-time validation layer for LLM applications.
- LLM EvaluatorsLLM Evaluators let you define a judge prompt.
- Continuous EvalsA tool that runs LLM-as-a-judge evaluations over incoming traces, also known as online evaluation.
- TransformsA transform is a named, reusable extraction configuration scoped to a task.
- Prompt LibraryA single source of truth for every prompt your AI application uses.
- Agent DiscoveryFinds agents by scanning your environment and adds them to one inventory.
- Alert RulesAn alert rule is a named SQL query that Arthur evaluates on a recurring schedule.
- Custom EvalsLets users define, configure and reuse their own performance and quality metrics.
- SSO (OIDC and SAML)Connect any OIDC 1.0 or SAML 2.0 compliant identity provider.
- RBACVersion, reuse and govern metrics across teams and projects with full RBAC and auditability.
- Data ConnectorsData Connectors let you point Arthur at your existing data warehouse or object storage.
- MCP server monitoringWatches for new MCP servers to catch agents as they come online, and flags capability changes in real time.
- Webhook templatingA webhook templating engine using Jinja2 syntax.
- Model fairness and bias detectionShows how your model treats different population groups, with fairness thresholds you set for your business.
- Guardrail threshold tuningMany guardrails offer custom configuration to set a per-use case threshold on where they trigger.
Sources
- Pricing
-
Three named tiers read live: Free $0/mo; Premium $60/mo; Enterprise custom.
- Description
-
Official documentation · 9 Sep 2026
Own homepage and pricing page name model monitoring, bias/drift detection and the open-source Arthur Evals Engine.
- Sold within
-
Bought and metered on its own pricing page -- no parent offering it ships inside.
- URL
-
Official documentation · 9 Sep 2026
Fetched 2026-09-09: markets as an independent AI observability product.
- Deployment
-
No self-hosted, VPC or on-premise option named at any tier, including Enterprise -- SaaS only, per the live pricing page.
- Interfaces
-
Platform listing · 14 Oct 2026
PyPI listing published by Arthur: 'Arthur Python API Client Library'.
- Intended for
-
Official documentation · 13 Oct 2026
Page: 'Perfect for small teams getting started with AI' / 'Advanced monitoring, SSO, SLAs and BAA'.
Also from Arthur AI
-
Arthur Engine Developer tool
Open-source, real-time evaluation engine for generative AI and traditional ML models: instant analysis of outputs, active guardrails, and built-in plus customizable metrics, run inside a user's own infrastructure independently of the hosted Arthur platform.
-
Arthur Agent Security & Governance Application
Module within the Arthur AI Platform providing security scanning, policy enforcement and governance controls specifically for autonomous agent deployments (tool-use permissions, action auditing, risk scoring).
-
Arthur Observability Platform
Monitors model performance, model metrics, and data drift.
-
Arthur Agent Development Toolkit Developer tool
Toolkit for building, debugging and shipping AI agents: version prompts, A/B test prompts, models and RAG configs on production data, trace every agent step, run continuous evals and guardrail checks for hallucination, PII, prompt injection and toxicity on live traffic, and alert on quality drift.
Open Arthur AI in the directory
Something wrong here? Send a correction — quote this product id: arthur-ai-platform.