Idea Launch MCP

Get your market read without leaving your editor

You are already in a coding agent when the question comes up: is this worth building? Paste one line and it installs the Idea Launch plugin itself. Then create the idea, draft the brief and the ad, and read the result — in Cursor, VS Code, Copilot, Codex, Claude Code, and more.

Start an idea test See how it works

Want to see the whole loop first? Watch a session go from install to market read.

Ask your agent to install it

You are already talking to an agent. Let it do the setup — copy this and paste it in:

Install this Agent Plugin: https://github.com/apsquared/idea-launch-plugin

Read the repo's README and install it the way my client does it. If this client cannot install plugins, connect its MCP server instead — https://www.idea-launch.io/api/mcp, streamable HTTP, no API key. Then tell me anything I still have to do myself, like a restart or a browser sign-in.

That is the whole install. We publish an Agent Plugins package — an open, vendor-neutral format — so your agent connects the Idea Launch server and adds two skills in one step. It works in Cursor, VS Code, Copilot, Codex, Claude Code, and more.

Two things to expect. Most clients need a restart before the plugin loads, and the server stays signed out until your first call opens a browser sign-in — until then it may report as failed to connect, which is the unauthenticated state rather than a broken install.

Or install it yourself

Every client installs plugins its own way. Point yours at the repository:

https://github.com/apsquared/idea-launch-plugin

Then follow your client’s own instructions:

In Claude Code, it is two commands:

/plugin marketplace add apsquared/idea-launch-plugin
/plugin install idea-launch@apsquared

More clients adopt the format over time — the current list is here.

What the skills add

The server gives your agent the tools. The skills give it the judgement to use them well — which is the difference between a test worth paying for and a wasted one.

Writing a brief worth targeting against: “Founders” is not a target customer. The skill pushes your agent to get specific, because the brief is what the targeting and the ad copy are built from — a vague one quietly wastes the run.
Reading the result honestly: A score of 45 at low confidence is not “mixed signal, refine the angle” — it is closer to “not settled yet”. The skill teaches your agent what the bands mean, when confidence makes a score unusable, and when a result cannot speak to demand at all.

Then just ask

Once it is installed, there are no commands to learn. Say what you want:

“I just built a scheduling tool for barbers. Set up an Idea Launch test for it.”
“Is my idea ready to test yet, or is something still missing?”
“How did my Idea Launch test go, and what should I do next?”
“Why isn't my ad run live yet?”

Your agent creates the idea, drafts the positioning for you to correct, generates the creative, then tells you exactly what is still missing before a test can run. When you are ready, it hands you the link to launch.

Read the FAQ

Or connect the server on its own

If your client does not support the plugin format yet, connect the MCP server directly. You get all the same tools — just without the two skills.

Claude Code

Run this once in your terminal.

claude mcp add --transport http idea-launch https://www.idea-launch.io/api/mcp

Codex

Add this to ~/.codex/config.toml.

[mcp_servers.idea-launch]
url = "https://www.idea-launch.io/api/mcp"

OpenCode

Add this to opencode.json.

{
  "mcp": {
    "idea-launch": {
      "type": "remote",
      "url": "https://www.idea-launch.io/api/mcp"
    }
  }
}

Any other MCP client? Point it at https://www.idea-launch.io/api/mcp over streamable HTTP.

There is no API key either way. The first call opens a browser sign-in, and your editor remembers it from there.

What your agent can read

Your ideas: List them, check setup progress, and see which are ready to test.
Your market read: The validation score, the signals behind it, and the recommended next move.
Your Ad Runs: Status, click-through rate, traffic, and signups for every run.
Run timelines: Why a run is not live yet, including a hold in our review.

What it can set up for you

Everything up to the point of spending is free and works from the agent.

Create an idea: Start a new test, or point us at a URL you have already shipped.
Draft the brief: Target customer, why they would care, the current alternative, the market.
Write the ad: Tagline, ad copy, and an image prompt, generated the same way the app does it.
Stage the test: Check every launch gate and preview the platform and audience that would run.

What it will never do

No tool on this server can spend your money. Starting an Ad Run uses one of your Ad Runs, and that is confirmed by you in the app — never by an agent. The server can tell you an idea is ready to test, show you the platform and audience it would use, and hand you the link. You click the button.

Your agent also only ever sees the same numbers the app shows you: click-through rate, landing-page traffic, signups, and the validation score. An Ad Run tells you whether people want this positioning — the message, the offer, and the page together. It is real evidence, not proof the business will work.