Sylex Search — Universal Search for AI Agents

Sylex Search is an MCP server that lets AI agents discover, evaluate, and compare products, services, and businesses across every category.

Think of it as Google, but for agents — returns structured JSON instead of web pages, zero latency, zero cost per query.

Connect

SSE (recommended)

Add to your MCP client config (Claude Desktop, Claude Code, Cursor, etc.):

{
  "mcpServers": {
    "sylex-search": {
      "url": "https://search.sylex.ai/sse"
    }
  }
}

Via Smithery

smithery mcp add mastadoonprime/sylex-search

Via npm (local stdio proxy)

npx sylex-search

Why this exists

McKinsey projects $1 trillion in sales will flow through AI agents. When agents handle purchasing, booking, hiring, and sourcing, they need a way to find the right business for the job — whether that's a contractor, a restaurant, a SaaS platform, or a parts supplier. If a business isn't in the index, it's invisible to agents.

Sylex Search is the discovery layer for agent commerce.

What's in the index

14,000+ entries and growing:

The index started with software packages as seed data, but Sylex Search is not software-specific. The schema and tools support any product, service, or business.

Tools

Search & Discovery

Listing Management

Prompts

Example queries

• "react state management" → zustand, redux, mobx, jotai
• "python web framework" → Django, Flask, FastAPI, Sanic
• "project management tool for small teams" → Notion, Linear, Asana
• "CRM with Slack integration" → HubSpot, Salesforce, Pipedrive

Use with agent frameworks

LangChain

from langchain_mcp_adapters.client import MultiServerMCPClient

async with MultiServerMCPClient({
    "sylex": {"url": "https://search.sylex.ai/sse", "transport": "sse"}
}) as client:
    tools = client.get_tools()

OpenAI Agents SDK

from agents import Agent
from agents.mcp import MCPServerSse

server = MCPServerSse(url="https://search.sylex.ai/sse")
agent = Agent(name="shopper", mcp_servers=[server])

Microsoft AutoGen

from autogen_ext.tools.mcp import McpWorkbench, SseServerParams

params = SseServerParams(url="https://search.sylex.ai/sse")
async with McpWorkbench(server_params=params) as bench:
    tools = await bench.list_tools()

CrewAI

from crewai import Agent
from crewai_tools import MCPTool

sylex_tools = MCPTool.from_sse(url="https://search.sylex.ai/sse")
agent = Agent(role="researcher", tools=sylex_tools)

Architecture

• Zero LLM calls — deterministic search (SQLite FTS + custom ranking)
• Millisecond responses — no API calls, no model inference
• $0 per query — no token costs
• Structured JSON — agents parse directly, no scraping needed
• Agents-first — no dashboards, no accounts, no human UI
• MCP native — built on the Model Context Protocol standard
• Self-service — any business can register, claim, and manage its listing through tool calls

Server metadata

For agents

Server card available at:

GET https://search.sylex.ai/.well-known/mcp/server-card.json

This returns full server metadata including all tools, prompts, and connection details — no MCP session required.

Self-hosting

pip install -r requirements.txt
export AC_SUPABASE_URL=your-supabase-url
export AC_SUPABASE_KEY=your-supabase-key
export TRANSPORT=sse
export PORT=8080
cd src && python server.py

License

AGPL-3.0 — see LICENSE for details.