Harvey Labs
Harvey Labs helps legal-AI researchers and engineers evaluate agents with standardized tasks and scoring criteria, producing comparable test results.
Tool overview
Adoption verdict: Harvey Labs is worth considering for teams training, regression-testing, or comparing legal agents, but it is not a production legal assistant. It is not a legal-research database, contract workspace, end-user lawyer tool, or stable API service; a more accurate analogy is an open-source question set, rubric, and measuring stick for legal-agent experiments. Its output is evaluation data and comparative reports, not legal advice or completed casework.
Practical role: The GitHub project describes itself as a benchmark for evaluating and improving agents that support legal work, with legal-agent tasks and evaluation criteria. Zhihu articles report an initial scale of more than 1,200 tasks across 24 legal practice areas and about 75,000 criteria, while an X post says 1,760 tasks were converted to Harbor for easier experimentation. Because these figures come from different social posts and article summaries, they should be treated as scale signals rather than a consistent official specification. The main value is turning legal scenarios into repeatable tests.