AI Launch Kit — the open-source website foundation for the agent-first internet

The Apache 2.0 open-source website starter built for AI search. Agent-ready, AEO-scored, deployable in days.

AI Launch Kit — the open-source website foundation for the agent-first internet

The Apache 2.0 open-source website starter built for AI search. Agent-ready, AEO-scored, deployable in days.

Built by the Working Theory team and used by the team to run Sam Henry, Working Theory, and SalesSmyth on the same substrate. Three paths below. Pick the one written for you.

Built to be cited, from day one

45

Page schema types

21

Section types

14

MCP tools

Apache 2.0

License

The page is the unit of citation, not the site

When an LLM recommends a business, it assembled the recommendation from individual pages it read, attributed to a brand, and trusted enough to quote. The brand that gets cited is the brand whose pages got read correctly. Sites without declared entities lose at the attribution step. The model cannot tell which entity a page belongs to, so it cites a competitor whose page made it obvious. AEO work fails silently at that point. Nothing in the toolchain reports it. That is why the foundation matters more than the chrome. The structural prerequisites for being citable sit below the visual layer: typed schema, declared entities, agent-readable surfaces. A site can look polished and still fail the citation step.

Four structural capabilities make a site citable at deploy time

Each capability is verifiable today — not a roadmap promise. Clone the repo and run ailk audit to see for yourself.

Typed schema registry

45 page types declare every entity correctly. The model knows what the page is about and which business it belongs to.

First-class MCP server

CI-enforced parity with the human UI. Agents researching on a buyer's behalf find the same surface, not a degraded approximation.

AEO scoring CLI

ailk audit runs the citation checks in CI before a site ships. Structural problems get caught in CI, not three months after launch.

Agent-assisted scaffolding

From reference brand to deployed site in days. The scaffolding skills target Claude Code agents with a reproducible install path.

Built so a model can read a page, identify the brand, and trust the structure enough to quote

AI Launch Kit is the open-source website foundation for the agent-first internet — Apache 2.0, agent-ready, AEO-scored, deployable in days. Four structural capabilities make a site citable at deploy time. A typed schema registry of 45 page types declares every entity correctly. A first-class MCP server holds CI-enforced parity with the human UI, so agents and humans read the same surface. An AEO scoring CLI runs the citation checks in CI before a site ever ships. Agent-assisted scaffolding takes a project from reference brand to deployed site in days. Built by the Working Theory team, who run their own brands on it. The substrate is the same one the maintainers ship from.

Choose your path

Three routes, one substrate. Each card lands you on the page written for your role.

briefcase

I run an AI-search agency

AILK is the substrate to ship SMB sites that work for agents and humans from day one. The schema registry, the MCP server, and the AEO scoring CLI are the structural pieces your audit findings already point at.

code

I am a developer or indie hacker

Apache 2.0 monorepo. pnpm create ailk@latest to scaffold; ailk audit to score; deploy on Vercel from a fresh clone. Typed everything, peer apps for web, API, and MCP, no lock-in.

users

An agency told me to look at this

AILK is the website foundation your agency uses. ChatGPT and Claude can read pages built on it. That is why they recommended it. The agency runs the project; you get a site that shows up when AI search recommends businesses like yours.

What teams say about shipping on AILK

We had a site that looked polished but never showed up in AI search answers. After deploying on AILK, ChatGPT started citing our content within two weeks. The schema registry was the piece we were missing.

Agency founder, Austin TX

AI search agency principal

The MCP server came out of the box. Our client's AI assistant can now answer questions about their services using the same data the website surfaces. That took a day to wire, not a sprint.

Principal engineer, SaaS startup

Full-stack engineer

Quick answers

Is this just another SaaS starter?

No. AILK is built for the agent-first internet. SaaS starters were built when Google was the only entry point, so they optimize for dashboards and auth. AILK optimizes for being read correctly by the systems now recommending businesses.

Do I need to know Next.js to use this?

A developer with React experience can scaffold a site in an afternoon. Non-developers work through an AI-search agency that runs the project. Either path lands on the same substrate.

What is free versus paid?

The Apache 2.0 OSS is the full substrate: every schema, the MCP server, the AEO scoring CLI, the scaffolding workflow. Paid tiers add capability; they never gate access.

Pick a path

Three routes from here. Pick the one written for you, or star AILK on GitHub and come back when you are ready.