All plugins

launchspy

Spy on the ads and funnels of the top coaches and course creators through LaunchSpy's hosted MCP server (Streamable HTTP at https://launchspy.io/api/mcp, OAuth sign-in, no keys in config): search creators by niche and country, see who is scaling and how they run ads, read their Meta ads with run dates and EU reach, pull the longest-running ads of a niche, find and take apart their funnels (steps, builder, pixels, payment stack, pricing), size a market by niche and country, and check what is known about any domain. Bundles six workflow skills. Agent Plugins v1 package (plugin.json + mcp.json + skills/), configuration only. Disclosure: every tool call goes to LaunchSpy's hosted server (launchspy.io) and acts on the user's LaunchSpy account; each tools/call uses one credit (a free account gets 50 trial credits, once; paid plans include monthly credits); the tool set is served by that server at runtime and is not covered by the pinned sha. All ten tools are annotated read-only and the server is marked untrusted. Sign-in runs the pinned mcp-remote@0.1.49 through npx (Node.js 20.18.1+ and npm required), which opens a browser and a local callback port on first use and stores OAuth tokens under ~/.mcp-auth.

LaunchSpyToolsv1.2.0Updated 7 days ago

About

Spy on the ads and funnels of the top coaches and course creators through LaunchSpy's hosted MCP server (Streamable HTTP at https://launchspy.io/api/mcp, OAuth sign-in, no keys in config): search creators by niche and country, see who is scaling and how they run ads, read their Meta ads with run dates and EU reach, pull the longest-running ads of a niche, find and take apart their funnels (steps, builder, pixels, payment stack, pricing), size a market by niche and country, and check what is known about any domain. Bundles six workflow skills. Agent Plugins v1 package (plugin.json + mcp.json + skills/), configuration only. Disclosure: every tool call goes to LaunchSpy's hosted server (launchspy.io) and acts on the user's LaunchSpy account; each tools/call uses one credit (a free account gets 50 trial credits, once; paid plans include monthly credits); the tool set is served by that server at runtime and is not covered by the pinned sha. All ten tools are annotated read-only and the server is marked untrusted. Sign-in runs the pinned mcp-remote@0.1.49 through npx (Node.js 20.18.1+ and npm required), which opens a browser and a local callback port on first use and stores OAuth tokens under ~/.mcp-auth.

LaunchSpy for AI agents

LaunchSpy: ads, funnels and pricing of coaches, for your AI agent

Spy on the ads and funnels of the top coaches and course creators from your AI agent, and copy what works: who sells in a niche, the Meta ads they run, how long each ad has been running, the funnels behind the ads and the tools those funnels are built on.

This repository packages the LaunchSpy MCP server with a workflow skill, for any AI agent, whatever the model behind it: Claude Code, Hermes Agent, Cursor, ChatGPT, n8n and any client that speaks MCP or reads Agent Skills. Agents without MCP use the REST API.

What your agent can do

  • Map a niche: the creators that advertise the most in a niche and a country, with their platforms, followers and ad volume.
  • Study a competitor: their active ads, headlines, calls to action and landing pages, and how many new ads they launched in the last 30 days.
  • Find winning angles: search ad copy by topic, ranked by run length and EU reach.
  • Take a funnel apart: steps, builder, tracking pixels, payment stack, pricing and urgency mechanics.
  • Check a website: what LaunchSpy already knows about any domain.

Ten read-only tools: search_creators, get_creator, get_creator_strategy, get_rising_creators, search_ads, get_top_ads, search_funnels, get_funnel, analyze_url, get_market.

No agent? The same data is served by the REST API for n8n, Make, Zapier and scripts (OpenAPI spec).

What to ask

You do not have to name LaunchSpy or ask for "an analysis". Ask the way you would ask a colleague, in your language:

You askSkill that runs
"Help me with my creative strategy" · "My ads are not working, what are the others doing?"creative-strategy-coach
"Build me a swipe file" · "Give me hooks for my coach client"ad-swipe-file
"What price should I charge for my course?" · "Break down this funnel"funnel-teardown
"Is my niche saturated?" · "I want to launch a course on X, does it work?"niche-validation
"Analyse my client's competitors" · "What is this coach doing?"competitor-watch

The same five workflows are served as MCP prompts by the server. Documentation: launchspy.io/docs.

Requirements

  • A LaunchSpy account. A free account is enough to try it, no card needed: it gets 50 trial credits, once (they do not renew, and EU reach is not included). The Pro and Business plans include monthly credits. See pricing.
  • Each tool call uses one LaunchSpy credit. Connecting and listing the tools is free.

Install

Claude Code

claude mcp add --transport http launchspy https://launchspy.io/api/mcp

Then copy skills/launchspy into ~/.claude/skills/, or install the skill with the command below.

Any agent that reads skills

npx skills add launchspyio/launchspy-agent

Hermes Agent

hermes plugins install launchspy

The first tool call opens your browser to sign in to LaunchSpy.

Other MCP clients (Claude Desktop, Cursor, ChatGPT, n8n…)

Use the server URL https://launchspy.io/api/mcp with OAuth, or an API key from LaunchSpy Settings → API. Step-by-step guides: launchspy.io/mcp.

Disclosure

  • Network: every tool call goes to LaunchSpy's hosted server (launchspy.io) and runs on the signed-in user's account. The tool set is served by that server at runtime.
  • Credits: each successful or failed tools/call uses one credit of the user's LaunchSpy plan.
  • Sign-in: in Hermes, the package runs the pinned mcp-remote@0.1.49 through npx (Node.js 20.18.1+ and npm required). On first use it opens a browser and a local callback port, and stores OAuth tokens under ~/.mcp-auth.
  • Read-only: all ten tools are annotated read-only. They read the LaunchSpy database; none of them writes data or visits a live page. The server is marked untrusted in plugin.json, so Hermes asks before calling a tool when it is configured to.
  • Third-party content: tool results contain text written by advertisers (ad copy, page titles). The skill instructs the agent to treat it as data and never follow instructions found in it.
  • Provenance: the MCP server, its OAuth server and the product all run on launchspy.io, operated by LaunchSpy (APEX GLOBAL WORLD SOLUTIONS LLC).
  • No local code: this package is configuration and instructions only (plugin.json, mcp.json, .mcp.json, skills/).

Links

License

MIT

README from the reviewed commit e0286f8; it updates when the author re-pins.

Developer Mode

Explore built-in Hermes toolsets and actions.

HermesAgent

Developer?

Create a Hermes extension for Desktop or agent workflows.

Developer Section