browser-use/browser-harness

This page introduces an external open-source repository. It is not an HDATF product.

What it does

Connects an LLM agent to a real browser through a single editable CDP websocket. When a helper is missing, the agent writes it into its local workspace, so reusable helpers build up over tasks while the core code stays protected.

How it helps ATF

A reference for Harness on letting an executor add its own reusable tools during a task while keeping core code off limits. Worth comparing with how Harness keeps state and memory while a task runs.

License

MIT Permissive. Commercial use and changes are allowed if the copyright notice is kept.

More in this category

• datawhalechina/hello-agents
  Hello-Agents is a free tutorial from the Datawhale community on building AI agents from scratch. It covers agent concepts, paradigms such as ReAct, Plan-and-Solve and Reflection, context engineering, memory, evaluation and Agentic RL training.
• TencentCloud/CubeSandbox
  CubeSandbox is a sandbox service for AI agents, built on RustVMM and KVM, that creates hardware-isolated sandboxes. It runs on one node or scales to multi-node clusters and is compatible with the E2B SDK.
• tinyhumansai/openhuman
  OpenHuman is a local-first agent harness. It stores memory as scored Markdown trees in SQLite, mirrored as an editable Obsidian vault, runs fleets of agents on durable graphs, and researches the user's data and the web.
• huggingface/ml-intern
  An archived Hugging Face project that is no longer maintained. It was an agent that researched, wrote and shipped ML code using the Hugging Face ecosystem, with access to documentation, papers, datasets and cloud compute.
• fathah/hermes-desktop
  A community-maintained desktop app for installing, configuring and chatting with Hermes Agent. It offers a GUI for chat, sessions, profiles, memory, skills, tools, scheduling and messaging gateways instead of managing the CLI by hand.

Only repositories in the ranked Trendshift lists are included, and the lists are used only to find candidates. We do not copy their ranks. Descriptions, licenses and star counts come from each GitHub repository. The notes are our own reading. We have not tested these projects, and a place on a trending list does not prove quality.