Your team already uses AI everywhere. Now own it.
BundleHub turns scattered personal AI setups into one governed company capability: agents, workflows, rules, and knowledge, versioned in one bundle, compiled to every coding agent, installed in any repo with one command.
$ npm i -g bundlehub && bundlehub install acme/engineering-coreWorks with Claude Code, Codex, Cursor, and any agent that reads AGENTS.md.
AI everywhere. Owned nowhere.
Every engineer has a private AI setup. It works for them. It doesn’t work for the company.
Fragmented per tool
Three copies of every standard. None agree for long.
Fragmented per person
The best setups stay personal. New hires start from zero.
Ungoverned
Nothing to version, assign, or revoke. No view of who runs what.
Unmeasured
No record of what is installed, stale, or drifted.
AI-native is not more AI.
An AI-native company doesn’t use AI harder. It owns AI the way it owns code: defined, versioned, reviewed, distributed, and measured. Agents do the repeatable work. Your team owns the decisions. Everything is auditable.
Defined
One versioned bundle instead of tribal prompts.
Owned
Roles and access, with every change audited.
Distributed
Every repo, every agent, one command.
Measured
Installs, versions, and drift, visible in one place.
From tribal knowledge to installed standard.
Compose
Author rules, agents, workflows, MCP configs, and knowledge in the console. Import your existing .claude/ or AGENTS.md, or copy a marketplace bundle.
Publish
Cut an immutable semver release with notes. Assign it to the whole org or to groups. Every change lands in the audit log.
Install
bundlehub install org/bundle compiles to each agent’s native format, gitignored by default. update and status keep every repo current and report drift.
Everything a company standard needs.
One bundle, every agent
Write your harness once. Adapters compile it natively for Claude Code, Codex, and Cursor, plus a generic AGENTS.md fallback for everything else.
Governance built in
Owners, admins, maintainers, and members. Per-bundle access by person or group, and an audit row for every change.
Versioned like software
Immutable semver releases with notes. Repos report drift, and terraform-style plan / apply makes every sync reviewable.
Knowledge that ships
Standards, architecture docs, and runbooks travel inside the bundle, so every agent answers from the same sources.
Secrets never stored
MCP configs carry environment variable names, never values. There is no field for secrets, so a leak can’t start here.
AI agents are users too
A built-in MCP server and --json on every command. Your coding agent can install, sync, and maintain its own harness.
One standard. Three kinds of users.
For platform leaders
Compose the standard, publish updates, see who has which version.
For engineers
One command in any repo. Nothing new in git status.
For coding agents
MCP tools, clean exit codes, --json. Your agent installs it itself.
The fast path is one command.
Install compiles the bundle to every agent you use, fences what it writes, and keeps it out of git status. Update and drift reporting come free.
Start from a proven bundle.
The marketplace is a curated catalog of rules, agents, workflows, and MCP configurations. Copy one into your org, adapt it to your standards, and publish it internally. It’s yours from the moment you copy it.
From first org to first install in under a day.
Create your org, compose or copy a bundle, and have your whole team and their agents on the same standard by tomorrow.
$ npm i -g bundlehub && bundlehub install acme/engineering-core