Freedom251 › AI Agent Observability Tools
MLflow Tracing
What MLflow Tracing is
MLflow Tracing records inputs, outputs, and metadata throughout requests made by LLM applications and agents, helping teams examine bugs and unexpected behavior. Its trace details bring together a span tree, latency for each span, inputs and outputs, tool calls, exceptions, token counts, cost breakdowns, feedback, and ground-truth annotations. Teams can add tracing automatically through integrations for more than 40 LLM and agent libraries and frameworks, or instrument code with the Python and TypeScript SDKs. Named integrations include OpenTelemetry, OpenAI, Anthropic, LangChain, LangGraph, LlamaIndex, DSPy, and Vercel AI, among others. The tool is OpenTelemetry-compatible and supports GenAI Semantic Conventions for trace export and ingestion. The free, open-source plan stores trace data on your own infrastructure and supports hybrid deployment. Its production tracing package, `mlflow-tracing`, has a documented footprint 95% smaller than the full `mlflow` package. For production deployments, the documentation recommends asynchronous logging and sampling. Custom span processors can mask sensitive data on the client before spans are sent to a backend.
Who it suits
MLflow Tracing suits teams building LLM applications or agents that need to inspect request steps, latency, tool calls, and token costs. It is also a fit for teams that want open-source tracing data hosted on their own infrastructure and integration options across multiple frameworks.
What it does well
- Automatic tracing integrations cover more than 40 libraries and frameworks.
- Trace views include span latency, tool calls, exceptions, and token counts.
- Python and TypeScript SDKs support manual instrumentation.
- OpenTelemetry-compatible, with GenAI Semantic Conventions support.
- Client-side span processors can mask sensitive data.
What to know first
- The TypeScript SDK does not work in frontend browsers or Edge Runtime.
- OTLP ingestion requires MLflow 3.6.0 or later and a SQL-based backend.
- Production server guidance recommends a reverse proxy or VPN for TLS and authentication.
Limits
The free plan is open source and hosts trace data on your own infrastructure. OTLP ingestion requires MLflow 3.6.0 or later and a SQL-based backend store; the TypeScript SDK is limited to Node.js backend environments, not frontend browsers or Edge Runtime.
Plans
| Plan | Billed and limits | Price |
|---|---|---|
| MLflow Tracing | Open source · trace data hosted on your own infrastructure | Free |
As mlflow.org lists them, read 30 Sept 2026. The maker's own page has the final word.
Getting started
- Open the MLflow Tracing documentation website.
- Choose a supported integration for automatic tracing, or use the Python or TypeScript SDK for manual instrumentation.
- For OTLP trace ingestion, use MLflow 3.6.0 or later with a SQL-based backend store.
- For production, follow the recommendations for asynchronous logging and sampling.
- For production server TLS and authentication, use a reverse proxy or VPN.
Verdict
Choose MLflow Tracing if you need free, open-source tracing with detailed spans and broad LLM framework integrations, while keeping trace data on your infrastructure. Look elsewhere if your TypeScript tracing must run in a browser or Edge Runtime, or if your OTLP ingestion setup does not meet the documented requirements.
Questions
How much does MLflow Tracing cost?
The listed MLflow Tracing plan costs nothing.
Is MLflow Tracing open source?
Yes. The plan is described as open source, and trace data is hosted on your own infrastructure.
Which platforms does it support?
The listed platforms are API, self-hosted, and web. The TypeScript SDK works in Node.js backend environments, not frontend browsers or Edge Runtime.
Which integrations are listed?
They include OpenTelemetry, OpenAI, Anthropic, LangChain, LangGraph, LlamaIndex, DSPy, and Vercel AI, among other frameworks, providers, gateways, and tools.
Can it trace requests automatically and manually?
Yes. Automatic tracing is listed for more than 40 LLM and AI agent libraries and frameworks, and Python and TypeScript SDKs provide manual tracing APIs.
What does a trace show?
Trace details include a span tree, per-span latency, inputs and outputs, tool calls, exception details, token counts, cost breakdowns, feedback, and ground-truth annotations.
Facts from MLflow Tracing's own pages
- Automatic instrumentation
- MLflow offers one-line automatic tracing integrations for supported LLM and agent libraries, including OpenAI, LangChain, and DSPy. · mlflow.org, 21 Oct 2026
- Automatic tracing
- The integrations page lists automatic tracing for more than 40 LLM and AI agent libraries and frameworks. · mlflow.org, 30 Sept 2026
- Browser limitation
- The TypeScript SDK works in Node.js backend environments and is not compatible with frontend browser or Edge Runtime environments. · mlflow.org, 30 Sept 2026
- Company founded
- Databricks says it was founded in 2013 by the creators of Apache Spark, Delta Lake, MLflow, and Unity Catalog. · databricks.com, 21 Oct 2026
- Company headquarters
- Databricks says it is headquartered in San Francisco. · databricks.com, 21 Oct 2026
- Dataset creation
- Production traces can be used to build evaluation datasets that retain details about components such as retrievers and tools. · mlflow.org, 21 Oct 2026
- Developer support
- The repository directs users to documentation, GitHub issues, a mailing list, and Slack for help and project discussions. · github.com, 21 Oct 2026
- Documentation
- MLflow documentation has separate sections for LLM and agent development, tracing, and evaluation, and for traditional machine-learning workflows. · mlflow.org, 21 Oct 2026
- Evaluation
- MLflow connects trace inspection with evaluation results to help assess application quality. · mlflow.org, 21 Oct 2026
- Human feedback
- Users can attach feedback to traces with metadata such as the user and timestamp. · mlflow.org, 21 Oct 2026
- Integrations
- The integrations directory organizes automatic tracing support across frameworks, coding agents, and model providers. · mlflow.org, 21 Oct 2026
- Licence
- The public MLflow repository identifies its licence as Apache-2.0. · github.com, 21 Oct 2026
- Lightweight SDK
- The production tracing package, `mlflow-tracing`, is documented as having a 95% smaller footprint than the full `mlflow` package. · mlflow.org, 30 Sept 2026
- Maker
- The MLflow project is developed at Databricks, according to TrustRadius. · trustradius.com, 21 Oct 2026
- Maker and community
- The MLflow site describes the project as backed by the Linux Foundation and committed to open source for more than five years. · mlflow.org, 30 Sept 2026
- Manual instrumentation
- Python developers can instrument custom functions with the `@mlflow.trace` decorator or create spans directly. · mlflow.org, 21 Oct 2026
- OpenTelemetry
- MLflow Tracing is compatible with OpenTelemetry and supports GenAI semantic conventions for trace export and ingestion. · mlflow.org, 21 Oct 2026
- OTLP ingestion limit
- OpenTelemetry trace ingestion requires MLflow 3.6.0 or later and a SQL-based backend store. · mlflow.org, 30 Sept 2026
- Pricing
- The MLflow Tracing page describes the open-source product as free and says trace data can be hosted on the user's infrastructure. · mlflow.org, 21 Oct 2026
- Production monitoring
- Tracing records latency, token usage, and quality metrics to help monitor deployed LLM applications and agents. · mlflow.org, 21 Oct 2026
- Production package
- The product documentation recommends the `mlflow-tracing` package for production deployments and says it has a 95% smaller footprint than the full MLflow package. · mlflow.org, 21 Oct 2026
- Production use
- The documentation describes tracing as production-ready and recommends asynchronous logging and sampling for production deployments. · mlflow.org, 30 Sept 2026
- Purpose
- MLflow Tracing captures inputs, outputs, and metadata from each step of LLM application and agent requests to help diagnose bugs and unexpected behavior. · mlflow.org, 30 Sept 2026
- Rating on G2
- No product rating or review count for MLflow Tracing is stated on the opened G2 MLflow marketplace page. · ai.g2.com, 21 Oct 2026
Where MLflow Tracing stands on Freedom251
- AI Agent Observability Tools#5 of 31
- AI LLM Observability Tools#4 of 23