Developer Tools·Standings
Best LLM Observability Tools in 2026
LLM observability tools help you watch language-model apps. This list compares LLM tracing, evaluations, prompt management, agent tracing, retrieval tracing and token cost tracking.
All 26 tools have a free plan. Compare paid starting prices, free trials, deployment choices and whether each tool has an API or open-source code. The board marks these signals so you can match a tool to your existing setup.
- 26 platforms
- 1 with a public API on this page
- 25 with a free tier
- Checked 23 Sept 2026
- 1 P
Promptfoopromptfoo.devOpen-source tool evaluates prompts across models and compares outputs in a web view.
- 2 WB
W&B Weavesite.wandb.aiObservability software for production AI agents maps runs, tools, and sub-agents.
- 3 OOpenLITopenlit.io
Open-source tool for LLM traces, reviews, prompt versions, and spending records with a free plan.
- 4 O
Opikcomet.comAgent observability platform that logs actions, visualizes traces, and runs evaluation workflows.
- 5 EA
Evidently AIevidentlyai.comOpen-source toolkit measures LLM, RAG, agent, and model quality with visual reports.
- 6 L
LangWatchlangwatch.aiAgent platform tracing steps, costs, tokens, errors, evaluations, and annotations.
- 7 G
Giskardgiskard.aiAI platform that identifies agent vulnerabilities and produces structured assessment reports.
- 8 L
Langfuselangfuse.comOpen-source tool records prompts, replies, tokens, speed, tools, and retrieval details.
- 9 LLunarylunary.ai
Lunary is a platform for developers building AI chatbots and other LLM-powered applications.
- 10 A
AgentOpsagentops.aiAI app observability platform tracks LLM calls, tools, agent interactions, and replayable runs.
- 11 TTruLenstrulens.org
Open-source tool traces agent steps with latency, tokens, cost, and evaluation metrics.
- 12 L
LangSmithlangchain.comAgent engineering platform with offline and online evaluations plus custom review methods.
- 13 B
Braintrustbraintrust.devAI tool inspects agent traces, tool calls, costs, speed, and output scores.
- 14 R
Ragasragas.ioOpen-source library evaluates LLM apps with metrics for RAG, agents, SQL, and other tasks.
- 15 FA
Fiddler AIfiddler.aiAI monitoring tool checks agents and models with 100-plus built-in or custom measures.
- 16 G
Galileogalileo.aiObservability and evaluation platform for GenAI and agent applications with guardrails.
- 17 MA
Maxim AIgetmaxim.aiAI agent tool simulates runs, scores results, monitors behavior, and manages prompt versions.
- 18 APArize Phoenixarize.com
Free tool captures AI calls, retrievals, tools, scores, and prompt changes for review.
- 19 CA
Confident AIconfident-ai.comAI quality platform for evaluations, production traces, metrics, and live quality alerts.
- 20 H
HoneyHivehoneyhive.aiAgent monitoring tool shows each step, runs reviews, watches traffic, and sends alerts.
- 21 H
Heliconehelicone.aiAI gateway and observability platform for routing, logging, caching, and debugging model calls.
- 22 L
Laminarlaminar.shOpen-source observability platform that traces agents, replays runs, and evaluates datasets.
- 23 RFRespan (formerly Keywords AI)respan.ai
AI gateway with observability, evaluations, security checks, fallbacks, and model routing.
- 24 AAArize AXarize.com
Agent improvement platform with issue detection, managed investigations, and swarm monitoring.
- 25 PAParea AIparea.ai
Parea describes itself as a platform for experiment tracking, observability, and human annotation to help teams ship LLM applications to production.
Places #1, #2, #3, #4, #5 on this board are open to sponsors. Take #1 · Other ways to advertise
How to choose
Start with the free plans
Every item on this list has a free plan. That gives you a way to compare tracing, evaluations, prompt management and cost tracking before choosing a paid level.
Six tools also list a free trial. Helicone lists a 7-day trial, while other trial terms vary by provider.
Check the paid starting price
Paid pricing differs widely. Opik lists paid plans from $19 per month, Arize AX from $50, Helicone from $79, Galileo from $100 per month when billed annually, Confident AI from $200 and Braintrust from $249.
Fiddler AI lists paid pricing from $0.00. Some tools on the list only show a free plan, so check the provider's current terms before budgeting.
Match the tracing features
The comparison covers LLM tracing, agent tracing and retrieval tracing, plus evaluations, prompt management and token cost tracking. Decide which of these you need to follow in one place.
A tool focused on model calls may not cover the same workflow as one that also follows agents or retrieval steps. Use the feature columns to narrow the list.
Choose where it runs
All 26 tools offer API and self-hosted deployment options. Twenty-four have a web option; nine support Linux, and four each support macOS and Windows.
If your team needs a desktop operating system, check the platform marks. Self-hosted and API options may fit teams that do not want a web-only setup.
Review access and code
A public API is one of the comparison signals, so check it when the tool must connect with other apps. Five of the 26 tools are open source.
Open-source status and API access are separate marks. A tool can offer one without the other, so review both before deciding.
Questions
Do all listed LLM observability tools have a free plan?
Yes. All 26 items list a free plan.
How many tools are open source?
Five of the 26 listed tools are open source.
Which platforms are available?
All 26 offer an API and self-hosted options. Twenty-four have web access, nine support Linux, and four each support macOS and Windows.
What features does the comparison cover?
It covers free plan, paid starting price, LLM tracing, evaluations, prompt management, agent tracing, retrieval tracing, token cost tracking and deployment options.
Are free trials available?
Six tools list a free trial. Helicone specifies a 7-day trial.
Some top positions on this board can be paid placements, standing where they were sold. The facts on each platform are not for sale. How we rank · Disclosure