Semantic Search

Find resources on Universal API using natural language queries. Powered by AI embeddings (AWS Bedrock Titan Text Embeddings v2).

How It Works

Every resource (MCP server, MCP tool, agent) is automatically embedded and indexed. When you search, your query is embedded and compared against all resources by meaning — not just keywords.

Example: Searching "send email" will find a tool called gmail_send even though the words don't match exactly.

Quick Start

# Search all resources
curl -s "https://api.universalapi.co/search?q=web+search" | jq

# Search only MCP servers
curl -s "https://api.universalapi.co/search?q=web+search&type=mcp" | jq

# Search individual MCP tools
curl -s "https://api.universalapi.co/search?q=send+email&type=mcp-tool" | jq

# Search agents
curl -s "https://api.universalapi.co/search?q=document+analysis&type=agent" | jq

Resource Types

Type	Description
all	Search everything (default)
mcp	MCP servers
mcp-tool	Individual tools within MCP servers
agent	AI agents

Response Format

{
  "data": {
    "results": [
      {
        "resourceId": "mcp-xxx",
        "resourceType": "mcp",
        "name": "serpapi",
        "description": "Search Google, News, and Reddit via SerpAPI",
        "slug": "snowtimber/serpapi",
        "score": 0.87
      }
    ],
    "query": "web search",
    "count": 5
  }
}

Results are ranked by similarity score (0-1, higher = more relevant).

Authentication

Search is available without authentication for public resources. With a Bearer token, private resources are also included.

Knowledge Search

To search across your uploaded files (not marketplace resources), use the Knowledge Search endpoint instead.

Next Steps

• Search API Reference — Parameters and response details