LangWatch
Choose LangWatch by events per workflow, rather than treating one customer conversation as one event.
- Starting price
- Growth EUR 29/core seat/month
- Free access
- 50,000 events/month and two users
- Billing unit
- Core seats, events and extended storage
- Platform
- Cloud dashboard with SDK and OpenTelemetry ingestion
- Verified
- 1 Nov 2026
Overview
What you provide
- Send authorised application traces, evaluation datasets and explicit success criteria through a supported integration.
What you get
- The workspace returns execution traces, evaluation scores, simulation replays and prompt experiment results.
Choosing this tool
A retrieval, LLM call, tool call and evaluation can each become billable events. Instrument a representative multi-step agent before estimating the free allowance. Phoenix is useful when a locally operated tracing platform is the priority; promptfoo fits repeatable command-line evaluations. LangWatch adds managed tracing, simulations and prompt work in one workspace, but the event definition makes long agent chains materially different from simple chat.
A decision worked through
For one Growth core seat and 500,000 monthly events, the published event calculation is EUR 29 + 3 × EUR 5 = EUR 44, before extended storage, tax or model costs. This assumes the stated 200,000-event allowance applies once to that workspace.
LangWatch alternatives for agent evaluation and production tracing
Compare LangWatch with Arize Phoenix, promptfoo for this workflow, including the billing unit and the work left to your team.
| Product | Price / unit | Free allowance | Input → output | Key limit |
|---|---|---|---|---|
| LangWatchRead profileOfficial source ↗ | Growth EUR 29/core seat/month | Developer includes 50,000 monthly events | Agent traces and datasets yield evaluation results. | Calls and evaluation steps can count as separate events. |
| Arize PhoenixRead profileOfficial source ↗ | ELv2 self-hosted software; operating costs separate | No listed self-hosted subscription charge | Instrumented spans become traces and experiment datasets. | ELv2 restricts offering substantial features as a managed service. |
| promptfooRead profileOfficial source ↗ | MIT CLI software; inference billed separately | No core CLI licence fee | Test cases and model configurations yield comparative evaluations. | Remote providers can receive prompts despite local orchestration. |
Setup & workflow
- Instrument one complete agent run and inspect its nested calls.
- Create a narrow evaluation dataset with known failure examples.
- Count resulting events and required retention before selecting seats.
Requirements & installation
- Connect a small test application and decide which trace fields may leave your environment.
Limits & review
- Event volume can exceed conversation volume. Free retention is fourteen days; longer retention and additional core seats change the bill.
Your part
- Review failure cases and judge criteria before using a score to approve a production change.
Sources & verification5
This profile is based on official sources, not a hands-on product test.
Vendor descriptions and demos document advertised features. Editorial guidance is based on these sources.
- Official LangWatch pricinglangwatch.ai
- Official LangWatch productlangwatch.ai
- LangWatch official referencelangwatch.ai
- Arize Phoenix official referencegithub.com
- promptfoo official referencegithub.com
