Harness Useful FAQ

Harness Useful is a family of four open-source, local tools for coding-agent harnesses such as Oh My Pi, PI, OpenCode, Claude Code and Codex. Harness Desktop is the window, Harness Skills the workflow, Harness Memory the shared project memory and Harness Dashboard the usage and cost meter. Each works on its own.

Start with Harness Desktop. Install it, then open Ecosystem and choose Run recommended setup. After you confirm each part, it installs Harness Dashboard, Harness Memory and Harness Skills, registers their MCP servers in each detected harness and adds the family hooks. You can also install any tool by itself.

Yes. Each tool has its own repository, installer and README, and works without the others. Harness Memory is a standalone MCP server, Harness Dashboard is a standalone app, and Harness Skills installs straight into your harness. They work best together, and Desktop connects them through Recommended setup.

Harness Desktop supports Oh My Pi and PI, with partial OpenCode support and chat previews for OpenCode, Claude Code and Codex. Harness Skills supports Oh My Pi, Claude Code, OpenCode and Codex. Harness Memory works with any MCP client. Harness Dashboard reads 13 tools, including Gemini CLI, GitHub Copilot, Zed, Cline, Roo Code, Kilo Code and Cursor.

Yes. All four tools are open source on GitHub under the GNU General Public License version 3. Harness Desktop, Harness Memory and Harness Dashboard are GPL-3.0-or-later, and Harness Skills is GPL-3.0-only. You can read the source, install the tools at no charge and follow each project’s releases on its GitHub repository.

No. Harness Desktop has no telemetry, no accounts and stores no tokens. Harness Dashboard never uploads your usage. Harness Memory opens no network connection unless you set up a team database. The tools make a few optional requests, such as signed update checks against GitHub, which you can turn off.

On your machine. Harness Memory keeps notes in a SQLite file, memory.sqlite3, in the harness-memory folder of your user data directory. Harness Dashboard keeps usage in a local SQLite database. Harness Desktop keeps hooks as JSON files in harness-hooks/hooks. Harness Skills is plugin files that your harness reads.

Harness Desktop runs on macOS (Apple Silicon and Intel), Linux and Windows. Harness Memory supports macOS, Linux on x86_64 and Windows. Harness Dashboard supports macOS, Linux on x86_64 and Windows. Harness Skills installs into your harness with a short command and needs Node.js 22 or newer.

Through MCP servers and hooks. Harness Memory and Harness Dashboard are MCP servers that your harness starts. Harness Desktop registers them in each detected harness and adds hooks that ask agents to search memory first, save what they learned and check limits before long work. Harness Skills adds the us-* workflow skills.

Install Harness Desktop with one command: on macOS and Linux run the curl install script, on Windows run the PowerShell install command. Both verify the release checksum. Then use Ecosystem and Run recommended setup to add the other tools. Each tool also has its own install command on its page.

Harness Desktop is a native desktop app that runs on top of the coding-agent CLI you already have installed. You chat with the agent in a window, browse its sessions and branches, review the changes it made and send comments back, and manage its configuration and extensions.

It supports Oh My Pi and Pi fully. Claude Code and Codex are supported with chat as a preview. OpenCode is partly supported, with chat as a preview. Other harness CLIs are planned. The app finds the CLI on your PATH and does not install harnesses.

On macOS and Linux run the curl install script from the README. On Windows run the PowerShell install command. Both verify the release checksum. You can also download an AppImage, deb, rpm, dmg, NSIS installer or MSI from the GitHub releases page.

No. It has no telemetry, no accounts and stores no tokens. Credential files are never read. The only requests the app makes itself are the signed update check and a lookup of the companion tools’ latest release numbers on GitHub, and you can turn both off.

Hooks are plain JSON files, one per hook, in the harness-hooks/hooks folder of your data directory. The app itself stores only preferences, window geometry, a few UI choices, your hook approvals and a local log file. Project notes belong to the separate Harness Memory server.

Yes. Harness Desktop is open source under the GPL-3.0-or-later license, copyright Maciej Spiechowicz. The source and releases are on GitHub.

Its Recommended setup installs Harness Dashboard, Harness Memory and Harness Skills after you confirm each part, and registers their MCP servers in each detected harness. It also adds family hooks that ask agents to search memory first and check limits with the dashboard before long work.

Harness Skills is a set of owned us-* skills for Oh My Pi, Claude Code, OpenCode and Codex. These harnesses discover the skill descriptions, and the model follows the relevant instructions with its own native tools and workers. It takes a request to reviewed, verified work.

Oh My Pi, Claude Code, OpenCode and Codex. Each has its own install section. Claude Code gets 21 namespaced skills and seven native agents. OpenCode v1.x and v2.x are supported. Codex needs the codex CLI. All adapters need Node.js 22 or newer.

Use the commands for your harness. Oh My Pi, Claude Code and Codex install a marketplace plugin with a short command. OpenCode, and the full Codex install, use a curl script on macOS and Linux or an irm script in PowerShell on Windows. Restart the harness afterwards.

Natural development requests go through us-run-workflow, which inspects the code and explains the selected lane. The skills are prompt instructions, not enforced stage transitions. In OMP, run /useful-skills workflow disabled in a repository to switch it off there, or invoke us-ignore-workflow for one request.

Yes. Claude Code, OpenCode and Codex get a separate local memory server with memory_status, memory_search, memory_save, graph_build and graph_query tools, with project-scoped private files. The us-use-harness-memory skill works with the separate Harness Memory MCP tools.

Yes. This repository is under GPL-3.0-only. Third-party portions keep their stated licenses, recorded in the repository’s attribution table. It has no custom workflow executor, model router or phase database.

Harness Memory is a stdio MCP server, shipped as one local binary called harness-memory. Agents such as Claude Code, Codex, OMP, PI and OpenCode search it at the start of a task and save what they learn, so the next session in the same repository starts better informed.

Any MCP client can start it. The README documents setup for Claude Code, Codex, OMP, PI and OpenCode. Each harness keeps its MCP servers in its own settings, so you add the harness-memory entry once per harness. Harness Desktop can register it for you.

On macOS and Linux run the curl install script from the README. On Windows run the PowerShell command. Both download the binary and checksums from the GitHub release and install nothing unless the SHA-256 matches. Then register it with each harness and check harness-memory --version.

In one SQLite file named memory.sqlite3, in the harness-memory folder of your user data directory. On Linux that is ~/.local/share/harness-memory, on macOS ~/Library/Application Support/harness-memory and on Windows %LOCALAPPDATA%\harness-memory. Set HARNESS_MEMORY_DIR to use another folder.

Not by default. Without a team database the server opens no network connection. The embedder downloads its search model once if you turn vector search on. What an agent reads from memory goes to its model provider like the rest of its context.

It is open source under GPL-3.0-or-later. By default each person keeps a local SQLite file. Team memory points every teammate’s server at one shared database, which you set up and configure yourself.

Harness Desktop registers the server in every detected harness while Project memory is on in its Settings, and adds hooks that make agents search first and save last. The us-use-harness-memory skill in Harness Skills works with its MCP tools.

Harness Dashboard is a local desktop app that reads the sessions your AI coding tools already write to disk. It stores everything in a local SQLite database and shows interactive charts of where your tokens and money go, by project, model, session and prompt.

Claude Code, Codex, Gemini CLI, GitHub Copilot CLI, GitHub Copilot Chat in VS Code, OpenCode, pi, omp, Zed, and the Cline, Roo Code and Kilo Code extensions. It also reads Cursor usage from cursor.com or a CSV export. That is 13 tools.

On macOS and Linux run the curl install script from the README. On Windows run the PowerShell command. The installer verifies the release checksums. You can also download a deb, rpm, dmg or msi from the GitHub releases page.

No. Everything runs locally and your usage is never uploaded. The app makes only a few outbound requests, such as the optional update check and plan-limit checks, and none of them sends your usage. Each can be switched off in Settings.

In a local SQLite database on your machine, or wherever you point it. The app data folder is readable by your account only. You can even point several machines at one SQLite file on iCloud Drive, Dropbox, OneDrive or a network share.

Yes. harness-dashboard mcp is a local MCP server over stdio. An agent can check what a session or branch cost, how a budget stands or how much of a plan limit is left, for example before a large refactor. It reads the database only.

Yes. Harness Dashboard is open source under the GPL-3.0-or-later license. Its costs are API-equivalent prices in USD, not a bill.