MLflow Tracing
ObservabilityApache-2.0
MLflow Tracing records the inputs, outputs and metadata of each intermediate step of an LLM or agent request so developers can find where a problem starts. It has Python, JavaScript and TypeScript quickstarts, one-line auto-tracing for libraries such as OpenAI, LangChain, LlamaIndex, DSPy and Pydantic AI, and a decorator for manual spans. It is compatible with OpenTelemetry, and the documentation says trace data is hosted on the developer's own infrastructure, with a managed MLflow option linked separately.
A lighter mlflow-tracing package is recommended for production.

Pinout
- MCP server: not checked yet
- HTTP API: not checked yet
- command line: supported, verified
- Python SDK: supported, verified
- TypeScript SDK: supported, verified
- OpenAPI description: not checked yet
| Interface | Status | Source we read | Read on |
|---|---|---|---|
| MCP server | Not checked yet | — | — |
| HTTP API | Not checked yet | — | — |
| Command line | Supported, verified | MLflow docs, GenAI Tracing, demo command (uvx mlflow demo) | |
| Python SDK | Supported, verified | MLflow docs, GenAI Tracing | |
| TypeScript SDK | Supported, verified | MLflow docs, GenAI Tracing | |
| OpenAPI description | Not checked yet | — | — |
Facts
| Purpose | Captures inputs, outputs and metadata for each intermediate step of a request. MLflow docs, GenAI Tracing, read . |
|---|---|
| Languages | Python, JavaScript or TypeScript, and OpenTelemetry-instrumented apps each have a quickstart. MLflow docs, GenAI Tracing, read . |
| Integrations | The page names OpenAI, LangChain, LlamaIndex, DSPy and Pydantic AI, and mentions Vercel AI. MLflow docs, GenAI Tracing, Integrations, read . |
| OpenTelemetry | The page says it is fully compatible with OpenTelemetry and supports GenAI Semantic Conventions for export and ingestion. MLflow docs, GenAI Tracing, read . |
| Auto-tracing | One line per library, for example mlflow.openai.autolog(). MLflow docs, GenAI Tracing, Auto-tracing, read . |
| Hosting | The page says trace data is hosted on the developer's own infrastructure and links Self-Hosting and Managed MLflow pages. MLflow docs, GenAI Tracing, read . |
| Price | The page says MLflow is open source and 100% free. MLflow docs, GenAI Tracing, read . |
| License | Apache-2.0 (LICENSE.txt, Databricks copyright 2018). Repository LICENSE.txt file, read . |
| Last commit | . GitHub API, latest commit on the default branch of mlflow/mlflow, read . |
Strengths
- Python, JavaScript and TypeScript quickstarts are documented.
- Traces follow OpenTelemetry GenAI conventions.
- The software is documented as free.
Pricing
As written on MLflow Tracing's own page on the read date. Prices change: check there before you pay.
Alternatives to MLflow Tracing
All observability tools for agents
OpenTelemetry-based observability platform with a Python SDK for LLM agents.

Open-source tracing, evaluation and monitoring for LLM apps and agents, by Comet.

OpenTelemetry-native tracing, evals and guardrails for AI agents, self-hostable.

Open-source observability and evals for AI agents, with OpenTelemetry tracing.
Last check: . Method: public repository and docs read automatically, page and day written down. Next check due within 90 days. How we check
Is this your tool? Write to hello@agent-tools.directory from an address on its domain to correct a pin or a fact.