What has shipped.

A pre-launch battle test, and the fixes it earned

AfterLaunch ran a full battle test across the product before opening the doors wider, exercising signup, onboarding, the Feed and the agent surface the way real use would. The fixes it surfaced shipped in the same wave.

Among them: a gap that could let a signup slip past the invitation gate is closed. An agent onboarding key that outlives its collection window is no longer handed out. Tools that change things over MCP now carry the same write rate limits as the REST API, so the two doors agree. And a move you have already shipped or archived can no longer be flipped back to pending by a stale link.

Your first growth moves arrive faster, and new places to learn

The wait between finishing your Growth Snapshot and seeing your first prepared moves is now much shorter: the first moves, each briefed with its evidence and a working draft, land about a minute in, with the fuller ranked batch following in the minutes after. Speed does not lower the bar; the early drop draws on the same prepared work as the rest.

Three new public surfaces shipped alongside it: per-agent setup guides at /agent-guides covering Claude Code, Cursor and friends, this changelog, and an expanded machine-readable corpus at /llms-full.txt so an AI agent can read the whole developer surface in one fetch.

Distribution wave 1: an MCP server an agent can pick up on its own

The AfterLaunch MCP server is now much easier for a coding agent to adopt without a human walking it through the setup. The connection instructions now read like a short guide to what AfterLaunch does and when to reach for it, and every tool and argument describes itself in plain growth, SEO and AI visibility terms so an agent can find the right one by searching.

Each tool now signals whether it only reads or can change something, which keeps a careful agent from guessing. A new Growth section can be dropped straight into a repository CLAUDE.md or AGENTS.md, served as raw markdown, along with a free tool that writes it in for you.

One correctness fix rode along: setting a loop autonomy rung is a Founder feature in the app, and it is now a Founder feature over the API too, so the two surfaces agree.

The act scope: a key can now close the loop, not just read it

Every key you mint in Settings, and every key an agent obtains through onboarding, can now do the work rather than only describe it. Alongside reading your Feed, a key can ship a move, skip it with a reason, redraft the body before it goes out, and archive it. It can also mint a checkout link when you want to upgrade.

These actions spend nothing, so they are granted by default. The on-demand scans and Feed top-ups still need the write scope, which AfterLaunch issues, so a key handed to an agent cannot spend money on its own.

The agent surface goes wide

AfterLaunch now opens to software. There is a public API and an MCP server your coding agent connects to, reads your ranked growth moves from, and acts on. The human sign-up path stays, but pasting a few lines into an agent is now a first-class way in.

An agent can even set AfterLaunch up for you. It submits your site and email, you approve with one click by email, and your free Growth Snapshot and trial start there. Nothing runs until you approve.

What you can do is bounded by credits rather than by tier walls: any account on an active trial can mint a key and run the on-demand AI Visibility scan. Founder at $149 a month is the only paid plan at launch, alongside the free trial.