Harness Manager

Harness Manager is a native Mac app for managing an AI coding stack in one place. Described as the App Store and control center for AI coding harnesses on Mac, it lets you discover, install, and update tools such as Claude Code, Codex, OpenCode, and Pi, automatically detects what is already installed, manages MCP servers and AI skills, checks versions and paths, and diagnoses broken installations or configuration issues. It is built for developers and anyone working with multiple AI coding tools who wants their tools, connections, and updates together where they can see them. The app gathers an entire AI coding stack into one workspace, with the stated goal of less setup to manage and more space to build. Harness Manager is free and open source. The problem it addresses is familiar to anyone working with modern AI coding tools. Each new harness, MCP server, or skill arrives with its own install path, its own version, and its own update mechanism, and keeping track of them spreads across terminals, bookmarks, and a growing pile of browser tabs. Harness Manager targets that sprawl directly. Instead of checking each tool separately, you get a single view of what is installed, what is running, and what needs an update. The product's own framing is explicit: your tools, connections, and updates, together, where you can see them. The site also positions it as a way to keep up with the ecosystem without keeping twenty tabs open, and as a way to get your stack sorted and get back to building. Discovery is the first job Harness Manager takes on. The Discover screen lets you browse harnesses, MCPs, and skills by popularity, moving from well-known tools like Claude Code and Codex to options you have not tried yet. The app presents tools with logos and install options, so adding the next piece of a workflow does not require hunting down a separate download page. Harness Manager works in the other direction as well: it finds supported tools and configuration already present on your Mac, so your existing stack appears in one view rather than needing to be re-entered. Tools shown in this context include Claude Code, Codex, Gemini CLI, Cursor, OpenCode, Warp, Antigravity, Antigravity IDE, T3 Code, Conductor, Superset, Paseo, cmux, Orca, Herdr, and Emdash. Update handling is deliberately conservative. The app shows what needs an update, lets you see installed versions, and lets you compare available updates. Before anything runs, you can review the command and decide when it runs. The phrase used on the site is "Your updates. Your call." — the point being that you decide when an update runs rather than having changes applied silently. Reviewing the command first gives you a chance to understand what will happen to your environment before anything changes. The same workspace surface shows what is installed and what is running, so update decisions can be made with a view of the current state of the stack. Beyond tools and updates, Harness Manager surfaces the rest of your setup in plain sight. It shows provider configuration, described on the site as local configuration signals; MCP servers, described as your tools' connections; skills, described as reusable instructions; and processes, showing what is running and where. Placing these details alongside the tools that use them means you can see how a connection or a skill relates to a particular harness instead of inspecting configuration files one at a time. The app also checks versions and paths and can diagnose broken installations or configuration issues, which is useful when a tool that used to work stops behaving as expected. Harness Manager also includes benchmarks and rankings for comparing models. The app presents 31 ranked collections and seven comparison metrics, with data by Modelgrep. Collections are grouped into themes. General covers Smartest, Coding, Agents, Fastest, Low latency, Cheapest, and Free. Development covers Design, UI components, Full-stack apps, Mobile apps, Tool calling, and Long-context reasoning. Reasoning and knowledge covers Reasoning, Math, Science, Writing, Instruction following, RAG, and SQL and analysis. Deployment covers Local, Open-source, Small and fast, Long context, Vision, and Uncensored. Creative and specialist covers Data visualization, SVG, Game development, 3D, and Roleplay. Individual metrics include intelligence, coding, agentic performance, design, speed, latency, and context. The site notes that rankings are a starting point and that results depend on your task. The Harness Briefing is a built-in news surface that goes beyond a changelog. It collects new harnesses, useful ideas, and the story behind releases, drawing on publisher news, community finds, and official releases. Named sources on the site include OpenAI, Google Developers, Simon Willison, Hacker News, and project releases. Every story links to its source, so the original material is always one step away, and articles can be read inside the app. For developers trying to follow a fast-moving ecosystem, this replaces the habit of monitoring many separate feeds and tabs with a single, sourced feed inside the same app that manages the tools being discussed. Getting started follows three steps. First, download, drag, and open: you move Harness Manager to Applications and open your workspace, with no build tools required. Second, see your existing stack: the app finds supported tools and configuration already on your Mac and shows them together in one view. Third, make your next move: review an update, discover a skill, or compare models for your next project. That sequence reflects the product's overall approach — begin from what you already have, make the current state visible, then act from that shared picture rather than from guesswork or scattered notes. It is a native Mac application, which is central to how it inspects local tools, configuration, and processes. The stated benefit is a stack that stays sorted. By concentrating discovery, installation, updates, configuration visibility, diagnostics, and model comparison in one native workspace, Harness Manager reduces the number of places a developer has to look and the amount of manual bookkeeping involved in maintaining a set of AI coding tools. Automatic detection means less time reconstructing what is already installed. Review-before-run updates mean changes to your environment remain under your control. Diagnostics for broken installations or configuration issues mean problems can be identified from inside the same app. And built-in model rankings and news mean evaluating new options does not require leaving the tool that manages your current ones. Concrete scenarios follow from these capabilities. A developer who has accumulated several coding harnesses can open the workspace to see everything installed alongside provider configuration, MCP servers, skills, and running processes in a single view. Someone preparing a new project can browse the Discover screen by popularity to find a harness, MCP, or skill and install it without leaving the app. A user who notices an available update can review the exact command before allowing it to run. When a tool stops working, versions, paths, and diagnostics can help identify a broken installation or configuration issue. Anyone weighing a model switch can compare options across the ranked collections — for example coding, agents, or tool calling — before committing. And a developer who wants to keep up with releases can read sourced stories from the Harness Briefing inside the app. Harness Manager targets Mac users who work with AI coding tools: developers, and anyone maintaining a collection of harnesses, MCP servers, and skills. It requires macOS 14 or later and runs on both Apple silicon and Intel. The site lists it as free to use, inspect, and make your own, licensed under Apache 2.0, with the project available on GitHub. It is distributed as a download for Mac rather than as a web service. The build is described as an early preview and is noted as not yet notarized, so macOS may require approval in Privacy & Security on first launch. No paid tier or pricing plan is mentioned; the product is presented as free and open source. In short, Harness Manager is a free, open-source Mac app that treats an AI coding stack as something to be managed rather than endured. It brings discovery, installation, updates, configuration and process visibility, diagnostics, model rankings, and ecosystem news into one native workspace built around what is already on your machine. The value proposition it states is simple: your stack, sorted, and back to building.