Straight answers about HubBound.

HubBound is the governance hub for AI tooling. Software teams version skills, rules, hooks, subagents and MCP servers as Artifacts, bundle them into Kits, distribute them to their teams and install them on every developer's machine with one command, then measure how AI tooling is used across the organization. It works like a private registry for AI agent configuration, with the organization layer public registries do not have: teams, roles, scoped distribution and organization-wide measurement.

An AI Artifact is a versioned, reusable unit of AI agent configuration in HubBound. It is one of exactly five types: a skill, a rule, a hook, a subagent or an MCP server. Every change to an Artifact's files or settings saves a new version, a History log records every event, and its visibility (Private, Organization or Public) can only be raised. An Artifact can be installed on its own or bundled into a Kit.

A Kit is a named, versioned bundle of AI Artifacts that HubBound installs and distributes as one unit. Each Artifact inside a Kit either follows its current version or stays pinned to one exact version, and every change to a Kit's contents saves a new Kit version. Instead of asking a new developer to copy nine files out of a wiki, you hand them one Kit and their environment matches the rest of the team.

The HubBound Marketplace is a read-only catalog where teams discover ready-made Kits and Artifacts: public ones published by the community and, on plans that include it, a private marketplace for their own organization. Each entry shows its version and its install command, and a team forks it into a private copy or distributes it to its teams instead of editing it in place.

Most teams start by pasting prompts into a wiki or copying rules files between machines, which drifts within weeks. HubBound replaces that with one governed source: skills, rules, hooks, subagents and MCP servers become versioned Artifacts, get bundled into Kits, and are distributed to specific teams or to the whole organization. A new version reaches every distribution that follows the latest version, pinned ones stay put, and the HubBound CLI and Runtime keep every machine in sync.

No. HubBound works alongside the AI coding tools your team already uses: it standardizes how they are configured, distributes those standards as Kits, and measures how they are used. Your developers keep their preferred tools.

HubBound works with five AI coding tools: Claude Code, Cursor, OpenAI Codex, GitHub Copilot and Antigravity. Kits are portable across all of them: the same Kit installs into any supported tool without rewriting the standards it contains.

HubBound is for engineering organizations of any size with more than one developer using AI: from startups standardizing AI usage from day one, to enterprises that need governance and visibility across many teams. The buyer is usually a head of engineering or platform lead; the daily user is the individual developer.

A HubBound organization has three roles: Org Admin, Team Admin and Member. Org Admins have full control, including Settings, members, billing and organization-wide distributions. Team Admins create teams, invite and manage their members, and distribute Kits and Artifacts to them. Members use what is distributed to them, with no admin screens to learn.

In HubBound, a Kit or a single Artifact reaches people through a distribution with three choices. The target: one team, or the whole organization, which only Org Admins can choose. The rule: Mandatory, which the HubBound Runtime installs for every member on its next background sync, or Optional, which each member decides. The version: always the latest, or pinned to one exact version until an admin changes it.

HubBound installs from its CLI page: download the full bundle (the HubBound CLI plus the HubBound Runtime) for Windows, macOS on Apple Silicon or Linux from the latest release, sign in with hubbound auth login, then install a Kit with hubbound install kit followed by its owner tag and name. From then on the Runtime syncs the machine's distributions in the background.

Public registries are directories: they help you discover an MCP server, but they do not know who is on your team. HubBound adds the organization layer: private Artifacts scoped to your company, Kits distributed to specific teams, three roles (Org Admin, Team Admin and Member), and AI usage measured across the organization. Discovery is table stakes; scoped distribution inside a company is the part public registries cannot do.

Artifacts and Kits that are not public never leave your organization. A Private one is visible only to its owner and the people given Read, Write or Admin access to it; an Organization one is visible to every member of your organization and to no one outside it. Visibility can only be raised, never lowered, and it is enforced by HubBound, not by convention.

Yes. HubBound Security audits the exact published content of a Skill version when a member asks for it: a deterministic ruleset flags known dangerous patterns such as piped remote scripts, credential access and undeclared network egress, and a model review reads for what rules cannot express, such as instructions addressed to the agent. The result is a verdict (clean, flagged, dangerous or inconclusive), a 0-100 risk score and every finding with its evidence and fix. The verdict is advisory: HubBound does not block downloads, installs or distribution on it.

Live Planning is HubBound's collaborative workspace for implementation plans, in Beta. The plan your coding agent writes becomes a shared Markdown document: your team edits it live and comments on selected passages, each member can launch their own coding agent (Claude Code, Codex, Cursor, GitHub Copilot or Antigravity, or any MCP client through the hubbound-live-planning server) to revise it, and reviewers approve a checkpoint before an agent executes the plan.

HubBound measures AI tooling on three levers. Adoption: sessions, active people, tools and models, and which Skills and MCP servers are used. Spend: tokens, cost by model and by person, cost traced to commits and pull requests, and prompt-cache savings. Delivery: commits, pull requests, their outcomes, time to merge and lines added, with the share that carries AI evidence. An Org Admin sees the whole organization or one member; everyone else sees their own activity. Metrics are not split by team.