MCP Server for AI Agents
Official Model Context Protocol server. Search launched products, weekly leaderboards, and open launch slots, or draft a launch for founder review from Claude, Cursor, or any MCP client.
Connect your agent to the live launch board
Point any MCP-compatible agent at our Streamable HTTP endpoint. Anonymous reading, verified owner reviews, and zero credentials shared with the model.
launchon.it/api/mcpPublic Tools
6
API Key
None
Access
Read + Draft
Transport
Streamable HTTP
Where to connect
Point your MCP client at the HTTP transport. Discovery files and server cards let crawlers auto-discover capabilities.
MCP Transport
POST JSON-RPC 2.0 messages. GET returns discovery info.
MCP Discovery
Streamable HTTP manifest describing tool capabilities.
Server Card
Single-file machine manifest with full schemas and transport metadata.
6 tools, zero authentication
Five read-only tools and one drafting tool. Typed schemas and parameter constraints match tools/list exactly.
search_products
Read OnlyKeyword search across every product that has launched on LaunchOn.it. Matches name, tagline and description, most upvoted first.
query (required) · limit 1-25 (default 10)get_weekly_leaderboard
Read OnlyRanked products for one ISO week, by community upvotes. Defaults to the live week, which closes Sunday 23:59 UTC.
week YYYY-Www (optional) · limit 1-50 (default 20)get_product
Read OnlyFull public listing for one product: tagline, description, who it's for, the problem it solves, pricing, tech stack, links and upvotes.
slug (required)get_winners
Read OnlyTop 3 podium finishers from recent weeks, newest week first. These are the Hall of Fame entries on /winners.
weeks 1-26 (default 4)get_launch_calendar
Read OnlyFree launch slots left in the live week and the weeks after it, plus the launch plans and prices. Use it to tell a maker when they would launch.
weeks 1-12 (default 4)submit_product
Drafts OnlyStarts a launch for a product URL. LaunchOn.it reads the site, drafts the listing (name, tagline, description, story, pricing, categories) and returns a private review link. The maker opens the link, signs in, edits and publishes. Nothing goes live and nothing is charged from this call.
url (required) · title · tagline · description · pricing_type (optional overrides)How submit_product works
Your AI assistant prepares the copy and metadata. You maintain absolute control over publishing and payment.
1. You ask your agent
"Launch my product https://example.com on LaunchOn.it." Your agent calls submit_product with the URL.
2. LaunchOn.it drafts it
LaunchOn.it reads your website, extracts copy, and drafts the name, tagline, description, story, pricing, and category tags.
3. You review & publish
The agent hands you a private review link. Sign in to your dashboard, fine-tune the story, pick a plan, and publish when ready.
Security Guarantee: An agent can never publish, charge a card, or pick a cohort week on your behalf. All drafts are saved privately, protected by 192-bit cryptographic review tokens, and automatically expire after 7 days if unclaimed.
Add LaunchOn.it to your AI client
No packages to install or build. Simply configure the remote Streamable HTTP endpoint in your tool of choice.
Claude Code
Run once in your project terminalclaude mcp add --transport http launchonit https://launchon.it/api/mcpClaude Desktop & Web
Settings > Connectors > Add custom connectorhttps://launchon.it/api/mcpCursor
Add to ~/.cursor/mcp.json{
"mcpServers": {
"launchonit": {
"url": "https://launchon.it/api/mcp"
}
}
}VS Code
Add to .vscode/mcp.json in workspace{
"servers": {
"launchonit": {
"type": "http",
"url": "https://launchon.it/api/mcp"
}
}
}JSON-RPC 2.0 over HTTP
Stateless request-response. Post one JSON-RPC message (or batch) per request and receive plain JSON in response.
{
"jsonrpc": "2.0",
"id": 1,
"method": "initialize",
"params": {
"protocolVersion": "2025-06-18",
"capabilities": {},
"clientInfo": {
"name": "my-assistant",
"version": "1.0.0"
}
}
}{
"jsonrpc": "2.0",
"id": 2,
"method": "tools/call",
"params": {
"name": "search_products",
"arguments": {
"query": "ai analytics",
"limit": 5
}
}
}curl -s https://launchon.it/api/mcp \
-H "Content-Type: application/json" \
-d '{"jsonrpc":"2.0","id":1,"method":"tools/call","params":{"name":"get_launch_calendar","arguments":{"weeks":4}}}'Other AI & developer surfaces
MCP is one entry point. Use the lightest surface that provides the information you need.
llms.txt
Compact site map and docs index for AI agents
llms-full.txt
Complete text of all live products, articles, and guidelines
ai.txt
Machine-readable permissions for AI crawlers
Markdown Twins
Clean markdown for any product: /products/{slug}.md
RSS Feed
Live feed of newly launched indie startups
Launch Rules
How weekly cohorts, community voting, and podiums work
Frequently asked questions
Do I need an API key?
No. Every tool works without an API key. Data queries are completely anonymous, and submit_product creates a private draft that you claim via magic link or OAuth in your dashboard, so no account credentials ever touch your agent.
Can an agent publish or pay for a launch?
No. submit_product only creates a private draft and returns a claim link. Selecting a plan, choosing a cohort week, and payment all happen securely in your browser once signed in. Unclaimed drafts remain completely private and expire in 7 days.
Should my agent scrape HTML instead?
Please don't. The MCP endpoint returns structured, typed JSON, and every public listing has a lightweight markdown twin at /products/{slug}.md. Both methods are significantly faster, use less bandwidth, and never break from DOM changes.
What rate limits are enforced?
60 requests per minute per IP for general queries, and 5 submit_product draft creations per hour. This is ample headroom for regular developer workflows and automated assistant tasks.
What tools should an agent call first?
Start by reading /llms.txt or calling get_launch_calendar to see current cohort availability. Use search_products and get_product to research competitors, or submit_product to draft a launch for review.