Alternatives to Arize Phoenix

Open-source LLM observability and agent evaluation you can trace, evaluate, and self-host for free.

Open-source LLM observability and agent evals: trace, evaluate, and improve production AI agents with MIT-licensed tooling.

OpenLIT is an open-source, OpenTelemetry-native agent harness engineering platform for LLM tracing, evals, and guardrails you self-host.

LangSmith is agent and LLM observability: trace, monitor, and evaluate AI agents in cloud, BYOC, or self-hosted.

TruLens traces AI agent runs with OpenTelemetry-native evals, showing where an agent fails and where you can cut cost.

Open-source AI observability for agent tracing, LLM-as-a-judge evals, and coding-agent cost tracking.

Developer observability platform that traces, replays, and debugs AI agent runs across OpenAI, CrewAI, Autogen and 400+ LLMs

Braintrust is an agent observability platform that traces AI runs, scores quality with evals, and surfaces production failure patterns.

Open-source agent observability that auto-captures every LLM and tool call and flags failures you never wrote a detector for.

Dash0 is OpenTelemetry-native observability where Agent0 investigates incidents, finds root cause, and opens fix PRs.

Opik by Comet turns LLM traces into root-cause groupings, eval scores and code fixes so agent teams stop shipping the same bug twice

Open source AI engineering platform for agent and LLM tracing, evaluation, prompt management, and ML lifecycle work.

Open-source AI coworker workspace that builds, schedules and ships agents to Slack or Telegram

Maxim AI evaluates, simulates and observes AI agents — with Bifrost, its open-source LLM gateway, handling routing, budgets and guardrails.

Open-source Python framework that replaces vibe checks with reproducible evaluation loops for RAG pipelines and AI agents.

Galileo is an AI observability and eval engineering platform that turns offline evals into production guardrails for agents and RAG.

Raindrop traces, detects, and simulates fixes for silent AI agent failures in production before they reach your users.

Metoro is a Kubernetes observability and AI SRE agent that watches your cluster with eBPF and opens fix PRs for production issues.

OpenJudge is an Apache 2.0 open-source AI evaluation framework shipping 50+ benchmark-validated graders for agents, LLMs, code, math, and multimodal models.

PandaProbe is open-source agent observability that traces runs, evaluates outcomes, and turns failures into validated fixes via its Repair

Galileo AI Evals turns offline evaluations into production guardrails that gate your AI agents in real time.

Open-source OpenTelemetry tracing, versioning, and sessions for Python LLM apps—bring your own backend.

LangChain builds open-source agent frameworks — langchain, langgraph and deepagents — plus LangSmith for agent observability, evals and

Athina AI is a collaborative AI development platform for managing prompts, running LLM evaluations, and monitoring production traces in one

Fiddler AI is an enterprise AI control plane for agent observability, inline guardrails, and auditable AI governance.

Weights & Biases tracks ML experiments and traces LLM apps so teams ship AI models faster

Evidently AI is an open-source Python framework for LLM evaluation, RAG testing, and ML model monitoring under Apache 2.0.

Local, MIT-licensed token and cost observability for AI coding agents — reads the logs your agents already write.

Free open-source AI usage tracking and analytics that syncs your token and cost data from Claude, Cursor, ChatGPT and other tools onto a shareable dashboard.

Silico is Goodfire's interpretability agent for understanding and debugging the internals of your AI models.