The long-horizon computer-use harness. Run AI agents across desktop apps and the CLI for extended periods while preserving task state and making reliable progress on complex workflows. Features fresh-context execution, durable verified state, independent auditing, recoverable progress, and native Claude Code / Codex / OpenClaw integration.
Install LongHorizon Harness
Installs but activates nothing: package.json has no `dsh.bundle` declaration, so `dsh plugin` warns and loads no config layer.
This repository is not a DSH plugin
package.json declares no dsh.bundle. In the official docs' own words, a package like this “still installs, but only as a plain dependency: dsh plugin prints a warning and activates no layer”. The install looks like it worked and nothing happens.What it actually is: The long-horizon computer-use harness. Run AI agents across desktop apps and the CLI for extended periods while preserving task state and making reliable progress on complex workflows. Features fresh-context execution, durable verified state, independent auditing, recoverable progress, and native Claude Code / Codex / OpenClaw integration. If that is what you came for, use it directly from the repository — it is just not something dsh plugin add can activate.
Before you run it
- A plugin runs with your permissions once loaded: it can read your files, use your credentials and reach the network. Tool approvals do not sandbox it.
- Installing from git resolves a moving branch. Pin a commit —
github:AMAP-ML/LongHorizon-Harness#<commit-sha>— so the code you reviewed is the code you install.
Prerequisites
- API key
- Required — see the repository README for which provider.
- Network access
- Required at runtime.
- Language
- Python
Install check
Every field the verdict was derived from, so you can re-derive it yourself
| Field checked | Result | Weight | What the spec says about it |
|---|---|---|---|
| lifecycleScripts | pass | info | no install-time scripts |
| dsh.bundle | missing | fatal | package.json has no `dsh.bundle` key |
- Verdict reason
- no dsh.bundle declaration — installs as a plain dependency only
- package.json read
- frontend/core/package.json
- Rules applied from
- docs/user/develop/basic/publish.md
- Checked at
- 2026-08-15
From the README
Excerpt as published by the author, plain text, unedited
**Give Claude Code, Codex, or DeepSeek Harness a goal once. Keep it working across desktop apps and the terminal for dozens of hours.** **Plan → act → verify → checkpoint or recover → repeat — until the work is actually done.** > **The model determines what an agent can do in one round. LongHorizon-Harness engineers the loop around it: what to do next, how to verify the result in the real computer, what progress to preserve, and how to continue after failure or context refresh.** **A Loop Engineering system for Claude Code, Codex, and DeepSeek Harness. One-command install, ready to run.** LongHorizon-Harness turns existing agents into long-running computer-use systems. Across desktop apps and the terminal CLI, it continuously recovers the goal and verified state, selects the next bounded s
Topics
GitHub topics on this repository
Related entries
Nearest entries by similarity — this one landed in no themed cluster
Dsh Plugin Greet
0lidaxiang
DeepSeek Harness is a plugin-based system for building AI agents. Everything, from tools and models to the web UI, can be added or replaced as a plugin.
Helloagents
hellowind777
一个自主的高级智能伙伴，不仅分析问题，更持续工作直到完成实现和验证。
Dsh Plugins
Wanbinyu
Community catalog of plugins and tools for DeepSeek Harness (dsh)
Dsh Plugin Session Delete
lsz-asd
Delete DeepSeek Harness sessions from the UI: header danger button + sidebar session-row menu item (no conversation jump), risk-consent dialog with session name/id, stops running agents first, in-place list refresh without page reload. Works in web and the desktop client.
Dsh Workflow
icetomoyo
把Claude Code的UltraCode模式带给DSH，把 DSH 的一次性多 Agent 调度，升级为可生成、可保存、可治理、可观察、可恢复的 Workflow 层
Dsh Genui
omdsh-dev
GenUI for DeepSeek Harness: interactive UI components rendered inline in assistant replies via the dsh-ui fence — layout, charts, plots, forms, quizzes, mermaid, 3D scenes, and an action event loop back to the model. Ships the fence-teaching host plugin, the browser renderer (client half), and the genui skill.