TruLens

Free plan Open source APISelf-hosted

What TruLens is

TruLens is an open-source Python SDK for tracing and evaluating AI agents, RAG applications, MCP apps, and summarization systems. It records per-step latency, inputs, outputs, token use, and cost, helping teams investigate poor responses. Evaluation metrics include tool selection, groundedness, context relevance, and correctness; users can also define measures with Python functions or LLM-as-judge prompts and tune judges using rubrics and examples. Evaluations can run as traces arrive or in batches over a dataset. A leaderboard compares app versions by metrics, latency, and cost. TruLens works with OpenTelemetry traces and can instrument custom Python apps and frameworks including LangGraph and LlamaIndex. Listed integrations and providers include LangChain, MCP, OpenAI, Anthropic, Amazon Bedrock, HuggingFace, LiteLLM, Google, and Snowflake Cortex. It logs to local SQLite by default or a SQLAlchemy-compatible database URL, and supports a Snowflake connector. The project began at TruEra and is now maintained as open source by Snowflake. The SDK is free and self-hosted, with installation through pip or from source. The Snowflake connector omits several conversation features, including grouping and dashboard views.

Who it suits

TruLens suits developers who want to trace and evaluate AI applications using Python and OpenTelemetry. It also offers options for teams using supported frameworks, providers, or Snowflake storage.

What it does well

Records step-level latency, inputs, outputs, tokens, and cost.
Supports real-time and batch evaluations.
Users can define custom metrics and judge prompts.
Free, open-source Python SDK.

What to know first

Snowflake connector lacks conversation grouping.
Conversation-level metrics are unavailable with the Snowflake connector.
Snowflake connector lacks retrieval APIs and dashboard views.

Limits

The Snowflake connector does not support conversation grouping, conversation-level metrics, retrieval APIs, or dashboard views. TruLens is self-hosted, so deployment and storage are managed by the user.

Plans

Plan	Billed and limits	Price
Open source	Open source Python SDK · install with pip or from source	Free
Open-source TruLens library	No subscription fee; users supply their own model/API and infrastructure costs · Python library · self-hosted storage/dashboard · optional provider packages	Free

As trulens.org lists them, read 7 Oct 2026. The maker's own page has the final word.

Getting started

Open the TruLens website
Install the Python SDK with pip or from source
Deploy TruLens in a self-hosted environment
Instrument a Python app or supported framework with OpenTelemetry
Choose local SQLite or a SQLAlchemy-compatible database URL for storage
Configure evaluations, metrics, and optional integrations

Verdict

TruLens offers tracing, evaluation, and version comparison for AI applications, with customizable metrics and self-hosted storage options. Teams relying on Snowflake connector conversation features should note the stated exclusions.

Questions

Is TruLens free?

Yes. The open-source Python SDK is listed as free.

Which platforms does TruLens support?

Its listed platforms are API and self-hosted deployment.

Can TruLens evaluate RAG and MCP applications?

Yes. Evaluation support includes agents, RAG applications, MCP apps, and summarization systems.

Can users create custom evaluation metrics?

Yes. Metrics can use Python functions or LLM-as-a-judge prompts, with rubrics and examples for tuning.

Where can TruLens store data?

It uses local SQLite by default, can connect through a SQLAlchemy-compatible database URL, and supports a Snowflake connector.

Who maintains TruLens?

The project was originally created by TruEra and is actively maintained as open source by Snowflake.

Facts from TruLens's own pages

Acquisition: Snowflake announced an agreement on May 22, 2024 to acquire TruEra's AI-observability platform. · snowflake.com, 15 Oct 2026
Agent benchmark: The homepage reports that Agent GPA caught 267 of 281 human-annotated errors, or 95%, on TRAIL and GAIA traces. · trulens.org, 15 Oct 2026
Agent metrics: Built-in agent metrics include tool selection, tool calling, tool quality, plan adherence, plan quality, execution efficiency and logical consistency. · trulens.org, 15 Oct 2026
API documentation: The project publishes a searchable API reference covering core classes, metrics, providers and connectors. · trulens.org, 15 Oct 2026
Batch evaluation: Runs can replay a dataset and compute the same metrics over stored inputs. · trulens.org, 15 Oct 2026
Company facts: LinkedIn lists TruEra as a privately held software-development company founded in 2019 with headquarters in Redwood City, California, and 51–200 employees. · linkedin.com, 15 Oct 2026
Current stewardship: The official TruLens homepage says the project was created by TruEra and is actively maintained in open source by Snowflake after the acquisition. · trulens.org, 15 Oct 2026
Custom evaluation: Developers can define metrics with Python functions, LLM-as-judge prompts, span attributes, rubrics and examples. · trulens.org, 15 Oct 2026
Custom metrics: Users can define custom metrics with Python functions or LLM-as-judge prompts and tune judges with rubrics and examples. · trulens.org, 7 Oct 2026
Dashboard: The built-in Streamlit dashboard provides leaderboards, trends, individual records and side-by-side version comparison. · trulens.org, 15 Oct 2026
Database connector: A Snowflake connector is provided for storing TruLens traces and evaluation data. · trulens.org, 15 Oct 2026
Evaluation: It supports agent, RAG, MCP app, and summarization evaluations with metrics including tool selection, groundedness, context relevance, and correctness. · trulens.org, 7 Oct 2026
Evaluation modes: Evaluations can run as traces arrive or in batch over a dataset. · trulens.org, 7 Oct 2026
Experiment comparison: The leaderboard compares app versions across metrics, latency, and cost. · trulens.org, 7 Oct 2026
Explained scores: Its judges attach explanations to metric scores so teams can investigate why a result was rated poorly. · trulens.org, 15 Oct 2026
Founders: Snowflake's announcement names Anupam Datta, Shayak Sen and Will Uppington as TruEra's three co-founders. · snowflake.com, 15 Oct 2026
Framework integrations: Official integrations instrument LangChain, LangGraph, LlamaIndex and NeMo Guardrails applications. · trulens.org, 15 Oct 2026
Independent company profile: Craft describes TruEra as an AI-observability company founded in 2019 and headquartered in Redwood City. · craft.co, 15 Oct 2026
Installation: The documented installation uses Conda with Python 3, followed by pip install trulens; source development uses the GitHub repository and editable installation. · trulens.org, 15 Oct 2026
Integrations: The documentation lists integrations and providers for LangChain, LangGraph, LlamaIndex, MCP, OpenAI, Anthropic, Amazon Bedrock, HuggingFace, LiteLLM, Google, and Snowflake Cortex. · trulens.org, 7 Oct 2026
Licence: The public GitHub repository identifies TruLens as MIT-licensed. · github.com, 15 Oct 2026
Limit: external evaluation costs: Evaluation can add model-token, infrastructure, database-retention and Snowflake costs, depending on deployment choices. · aicoolies.com, 15 Oct 2026
Limit: feature depth and scale: Conversion Gems characterizes the product as having fewer features than some alternatives, less depth and possible slowness at scale. · conversiongems.com, 15 Oct 2026
Limit: operational ownership: The independent review says production users must manage database lifecycle, access control, retention, upgrades and evaluator-worker reliability. · aicoolies.com, 15 Oct 2026