

TruLens
β Text source-checked Nov 13, 2026Open source library for evaluating and tracing AI agents.
Source-checked against the vendor's site Β· rebuilt from the vendor's own pagesProfile updated Link checked How we rate
In short
TruLens is a free, open-source library for evaluating and tracing AI agents. It uses LLM judges to score steps and identify failures or cost-saving opportunities.
TruLens is an open source library for evaluating and tracing AI agents. The vendor says that TruLens instruments applications with OpenTelemetry and scores every step with benchmarked LLM judges. TruLens identifies where an agent fails and where costs can be cut without losing quality.
TruLens records latency, inputs, outputs, tokens, and cost per step to provide a traceable cause for bad answers. TruLens judges explain their scores so that a low score points to exactly what went wrong. TruLens judges are graded against human annotations out of the box.
TruLens allows users to instrument any app in plain Python or auto-instrument specific frameworks. TruLens traces the user's app, judges with the user's chosen model, and writes to the user's database.
TruLens was originally created by TruEra. Snowflake actively maintains TruLens in open source following the acquisition.
According to TruLens: TruLens is an open source library for evaluating and tracing AI agents. Its pages list this strength: TruLens is open source and OpenTelemetry-native.
β The AI Tool Bible editorial team
Pros
- β TruLens is open source and OpenTelemetry-native.
- β TruLens judges are benchmarked against human annotations out of the box.
- β TruLens allows users to choose their own models and databases.
- β TruLens supports instrumentation in plain Python or via framework auto-instrumentation.
- β TruLens is actively maintained by Snowflake.
Facts
- Platforms
- api, cli, web (not source-checked)
- Pricing
- free (not source-checked)
- API available
- Yes (not source-checked)
- Open source
- Yes (license: MIT)
- GitHub stars
- 3,639 (checked Nov 13, 2026) (not source-checked)
- Last GitHub push
- Nov 12, 2026 (not source-checked)
- First commit
- Nov 2020 (not source-checked)
- Company
- TruEra (not source-checked)
- Model used
- Multi-model (LLM-as-judge) (not source-checked)
- Package downloads (30 d)
- 24,151 Β· pypi trulens Β· checked Nov 22, 2026 (not source-checked)
- Hacker News
- 1 story Β· top 4 points Β· last Aug 10, 2021 Β· checked Nov 22, 2026
- Earliest Wayback capture
- 2021-08-10 (earliest archived capture of the site, not a launch date)
Where this tool sits
Use cases
Frequently asked
- How much does TruLens cost?
- TruLens is free. It is an open-source library maintained by Snowflake following the acquisition of TruEra. No specific pricing details are documented beyond its free status.
- What models does TruLens use for evaluation?
- TruLens uses a multi-model approach with LLM-as-judge. It allows users to judge with their chosen model. These judges are graded against human annotations out of the box to ensure reliability.
- How do I integrate TruLens into my application?
- You can instrument any app in plain Python or auto-instrument specific frameworks. TruLens records latency, inputs, outputs, tokens, and cost per step, writing data to your chosen database.
- What specific metrics does TruLens track?
- TruLens records latency, inputs, outputs, tokens, and cost per step. It provides a traceable cause for bad answers and helps identify where costs can be cut without losing quality.
- Who maintains TruLens?
- TruLens was originally created by TruEra. Snowflake actively maintains TruLens in open source following the acquisition. It is designed for evaluating and tracing AI agents.