convez.ai
MCP Server
SplitMyGear

mcp-server-splitmygear

SplitMyGear/mcp-server-splitmygear

Ver no GitHub
InicialPluginScore · de 0 a 100

MCP Server for SplitMyGear — search, book, and browse outdoor gear rentals via AI agents

Instalação

$npx convez-ai SplitMyGear/mcp-server-splitmygear

1 comando no terminal: detecta o formato do repositório (marketplace ou plugin avulso) e instala via Claude Code. Requer Node 18+ e o Claude Code instalado.

Métricas

Estrelas
0
Forks
0
Acompanhando
0
Saúde
37%
Último commit
há 2 dias
Versão
v2.2.1
Issues abertas
1
Contribuidores
1

README

A tradução para o português deste README ainda está na fila.

mcp-server-splitmygear

MCP (Model Context Protocol) server for SplitMyGear — an outdoor gear rental marketplace. Lets AI agents (Claude, Cursor, Windsurf, etc.) search gear, check availability, manage bookings, browse experiences, and more.

18 tools across search, booking, pricing, content generation, experiences, and messaging.

Independent project — not affiliated with any third party.


Quick Start

Claude Desktop

Authentication is required. Every request must present either an API key (x-api-key) or a user JWT (Authorization: Bearer …). There is no public tier — the server wields privileged credentials, so unauthenticated requests are rejected with 401.

Add to your claude_desktop_config.json (API key required):

{
  "mcpServers": {
    "splitmygear": {
      "url": "https://mcp-server-splitmygear.vercel.app/api/mcp",
      "headers": {
        "x-api-key": "your-splitmygear-api-key"
      }
    }
  }
}

Cursor / Windsurf / mcp.json

{
  "mcpServers": {
    "splitmygear": {
      "url": "https://mcp-server-splitmygear.vercel.app/api/mcp",
      "headers": {
        "x-api-key": "your-splitmygear-api-key"
      }
    }
  }
}

Claude Desktop Extension (DXT)

Install manifest.json through Claude Desktop's Extension Manager (Claude Desktop v0.10.0+). The UI exposes the required API key field.

HTTP (direct)

curl -X POST https://mcp-server-splitmygear.vercel.app/api/mcp \
  -H "Content-Type: application/json" \
  -d '{
    "jsonrpc": "2.0",
    "id": 1,
    "method": "tools/call",
    "params": {
      "name": "search_listings",
      "arguments": { "query": "tent for a family camping trip this weekend" }
    }
  }'

Authentication

All access requires authentication — the server holds privileged platform credentials, so there is no public/anonymous tier. Present either:

  • an operator API key via the x-api-key header (unlocks read/search, pricing, and content tools), or
  • a user JWT via Authorization: Bearer <token> (additionally unlocks user-scoped tools — booking, messaging, personalized recommendations — which act as the authenticated user; they never accept a caller-supplied user id).

Unauthenticated requests are rejected with 401. The two tiers below mean "works with an API key" (Auth) vs "requires a user JWT" (User).


Available Tools (18)

Search & Listings

ToolAuthDescription
search_listingsAuthSearch gear by filters or natural language
get_listing_detailsAuthFull details for a listing
check_availabilityAuthCheck if a listing is available for dates
get_similar_listingsAuthSemantically similar gear
get_personalized_recommendationsUserRecommendations based on booking history

Bookings

ToolAuthDescription
create_bookingUserCreate a rental booking (Stripe payment)
cancel_bookingUserCancel a booking with optional refund
get_booking_statusUserStatus of YOUR booking (owner-checked)

Pricing & Business Intelligence

ToolAuthDescription
suggest_listing_priceAuthAI-powered price suggestion for a gear category
analyze_competitor_pricingAuthCompare a listing against local competitors

Content Generation

ToolAuthDescription
generate_listing_descriptionAuthAI description from name + keywords
improve_listing_titleAuthSEO-optimized title suggestions

Experiences

ToolAuthDescription
search_experiencesAuthBrowse outdoor adventures and tours
get_experience_detailsAuthFull info and available schedules
book_experienceUserBook spots on a scheduled experience

Messaging

ToolAuthDescription
send_messageUserSend a message to a renter or vendor
get_conversationsUserList your active conversations
generate_ai_message_draftAuthAI-drafted professional message

Natural Language Search

The search_listings query parameter accepts plain English. The server parses it into structured filters automatically:

"lightweight tent for solo backpacking next weekend under $40/day"
"water sports gear for 3 people in Seattle"
"climbing harness and helmet for a beginner"

Resources

Resource URIDescription
splitmygear://categoriesList of all 8 gear categories

Running Locally

Prerequisites

  • Node.js 18+
  • An operator MCP_API_KEY (and, for user-scoped tools, a backend JWT)
  • Access to the SplitMyGear backend REST API (defaults to production; override with BACKEND_API_URL)
  • (optional) An AI provider key (OpenCode Zen / OpenRouter / OpenAI) for the two content tools and the AI message draft

This server holds no Supabase or Stripe credentials — every action goes through the backend REST API, which is the single authority for auth, data, pricing and payments (SPLIT-226).

Setup

git clone https://github.com/SplitMyGear/mcp-server-splitmygear
cd mcp-server-splitmygear
npm install
cp .env.example .env.local
# Fill in your env vars
npm run dev

Your local server will be at http://localhost:3000/api/mcp.

Environment Variables

MCP_API_KEY=               # REQUIRED — operator key clients send via x-api-key
BACKEND_API_URL=           # optional — defaults to the production backend /api/v1
MCP_BACKEND_JWT_SECRET=    # optional — verify forwarded JWT signatures locally (defense in depth)
# AI provider for the content/draft tools (first key present wins): OPENCODE → OPENROUTER → OPENAI
OPENCODE_API_KEY=          # https://opencode.ai/zen — free tier
OPENROUTER_API_KEY=        # https://openrouter.ai — free tier
AI_CHAT_MODEL=             # optional override
AI_EMBEDDING_MODEL=        # optional override
MCP_RATE_LIMIT_TIER=public # internal | beta | public | default

Without MCP_API_KEY the server fails closed (every request 401s). Without an AI provider key the three AI content tools degrade gracefully (they return a "disabled" notice); every other tool is unaffected.

Tests

npm test

Deployment (Vercel)

This server is optimized for Vercel Serverless Functions. Push to the connected GitHub repo to trigger a deploy:

git push origin main

Rate Limits

TierRequests/minTool calls/min
public20200
beta50500
internal1001000

Set MCP_RATE_LIMIT_TIER in your environment to control the limit.


License

MIT

Ver o original no GitHub

Bom para

Relacionados