Long‑Horizon Computer‑Use Harness
The LongHorizon‑Harness lets AI agents work on desktop apps and the command line for a long time. It keeps the task state safe. The tool can pick up where it left off after a crash. Fresh‑context execution avoids stale data. Verified state ensures that each step is correct. Independent auditing records every action for later review.
The harness supports Claude Code, Codex, and OpenClaw out of the box. Integration is simple – just import the library and start the agent.
from longhorizon_harness import Harness
h = Harness(model=claude-code)
h.run(open spreadsheet, resume=True)
The project aims at complex workflows that need reliable progress. It can manage multiple apps, switch between GUI and CLI, and preserve data across sessions.
Learn more on GitHub: https://github.com/AMAP-ML/LongHorizon-Harness
Top comments (0)