Probe vs MLflow
MLflow is an open-source platform for the machine learning lifecycle: experiment tracking, a model registry and deployment tools, which you run yourself or use as a managed service. Probe is a hosted experiment tracker built for teams whose coding agents run the experiments.
Tell your coding agent:
Install Probe on this machine and its coding agents. Read https://research.prbe.ai/install.md and follow it.
How they differ
| Probe | MLflow | |
|---|---|---|
| What it is | A hosted experiment record for ML teams and their coding agents | An open-source ML lifecycle platform: tracking, model registry, deployment |
| Setup | Your agent signs you up and installs Probe; you sign in once in the browser | You run a tracking server or use a managed MLflow service |
| Logging from code | probe.init() and probe.log() | MLflow's tracking API |
| Coding agents | Work is recorded as it happens, by the Probe daemon or the agent itself; agents read the team's record with probe ask or a hosted read-only MCP server | MLflow's Claude Code integration traces the agent's own sessions (prompts, tool calls, tokens and cost); its MCP server lets an agent search traces and read experiments and runs |
MLflow facts above are from the MLflow documentation, its page on tracing Claude Code and its MCP server page, checked November 2026.
Pick MLflow if
- You want open-source software you operate yourself.
- You need its model registry and deployment tools in the same system.
- You want traces of what your coding agent did in each session.
Pick Probe if
- Your coding agents run most of your experiments, and you want each result linked to the agent session and code behind it.
- You want the record set up for you, with nothing to host.
- You track RL rollouts and want them linked to the runs and the agent sessions behind them.