ENVIRONMENT-DRIVEN AGENT DEVELOPMENT

Share tasks, benchmarks, leaderboards, and trajectories. Scale to billions of rollouts.

uv tool install harbor

Explore datasets on the hub

terminal-bench

Terminal-Bench is a benchmark for measuring agents' abilities to complete tasks using a terminal.

66 tasks

terminal-bench-science

A benchmark for evaluating AI agents on research workflows across the life, physical, earth, mathematical, and engineering sciences.

70 tasks

terminal-bench

terminal-bench-2-1

Version 2.1 of Terminal-Bench, a benchmark for evaluating agents in terminal environments.

89 tasks

datacurve

deep-swe-1-1

DeepSWE: Measuring frontier coding agents on original, long-horizon engineering tasks

113 tasks

abundant

swe-marathon

Ultra long-horizon software engineering tasks from SWE-Marathon.

20 tasks

scale-ai

swe-atlas-qna

SWE-Atlas - Codebase QnA is a benchmark of deep codebase comprehension and QnA problems for coding agents. Checkout https://github.com/scaleapi/SWE-Atlas/ for instructions on running it.

124 tasks

Discover more datasets

Publish tasks and datasets

Publish and share your tasks privately or publicly so they can be run with a single command.

npx skills add harbor-framework/harbor --skill publish

Sweep your benchmark in a single command

Launch from the CLI or directly in the Hub.

harbor run -c config.yaml --launch

Launch a job

Upload and share results

Bring local runs to the Hub. Explore trials and trajectories, then share your results with others.

harbor upload ./jobs/my-run

View results

Create leaderboards

Create custom leaderboards with Harbor Hub trials as receipts.

harbor hub leaderboard init \
  --package org/name \
  --name main \
  --output leaderboard.yaml
harbor hub leaderboard create --config leaderboard.yaml

View a leaderboard