AI AGENTS
harbor
by Harbor Framework
agentlead-sourcegithub-agent-framework-repositoriessource:github.comevaluationrl-environmentssandboxed-tasksopen-sourcegithub
Overview
Framework for specifying sandboxed agent tasks for evaluation and optimization.
Details
Harbor is an open-source framework from Harbor Framework for specifying sandboxed agent tasks for evaluation and optimization. The project is described by its GitHub repository and organization as a framework for running agent evaluations and for creating and using RL environments, with the official homepage positioned around sandboxed agent tasks.
When to Use
Use Harbor when you need to define sandboxed tasks for evaluating agents. Use it when you are working with agent evaluations that may connect to RL environments. Use it to explore optimization workflows around specified agent tasks.
Getting Started
- Visit the official Harbor homepage at https://harborframework.com/.
- Review the GitHub repository at https://github.com/harbor-framework/harbor.
- Inspect the repository examples or project documentation before using it in a production evaluation workflow.
Key Features
- •Specifies sandboxed agent tasks
- •Supports agent evaluation workflows
- •Supports creating and using RL environments
- •Official GitHub organization links to harborframework.com
Capabilities
- •agent-evaluation
- •sandboxed-tasks
- •rl-environments
Last updated Jun 6, 2026