# LLM observability & evals: the tools companies use

153 companies on usedby use at least one of the 6 tools in this category with a public customer. LangSmith leads with 43 companies. Counts are of companies on usedby, not of the whole market. Latest check 2026-10-14: 149 of 160 public lines checked.

> Tools that trace, test, evaluate and guard LLM apps and AI agents in production.

Family: [AI infrastructure](https://www.usedby.ai/tools/category/ai-infrastructure.md)

## Ranking

Public customers count every public line, job-post signals included; ties share a rank.

| Rank | Tool | Public customers | Confirmed | Observed | Signals |
|---|---|---|---|---|---|
| 1 | [LangSmith](https://www.usedby.ai/tools/langsmith.md) | 43 | 0 | 42 | 1 |
| 2 | [Braintrust](https://www.usedby.ai/tools/braintrust.md) | 42 | 12 | 29 | 1 |
| 3 | [Arize AI](https://www.usedby.ai/tools/arize-ai.md) | 35 | 7 | 28 | 0 |
| 4 | [Comet Opik](https://www.usedby.ai/tools/comet-opik.md) | 17 | 10 | 7 | 0 |
| 5 | [Galileo](https://www.usedby.ai/tools/galileo.md) | 13 | 1 | 12 | 0 |
| 6 | [Fiddler AI](https://www.usedby.ai/tools/fiddler-ai.md) | 10 | 1 | 9 | 0 |

---
Source: https://www.usedby.ai/tools/category/llm-observability-evals · How we check: https://www.usedby.ai/methodology