search.sh

The agent-native search engine.

Zero ads. One query. Infinite answers.

$ curl -fsSL https://search.prim.sh/install.sh | sh

Web search

News search

URL extract

llms.txt ● Mainnet

MCPToolsCLIRESTSDK

What agents use it for

Real-time research

Agents need current facts. search.sh returns clean, structured results — no HTML parsing, no ad noise, no rate-limit negotiation.

News monitoring

Market agents, alert systems, and news digests query the news endpoint on a schedule. Get headlines and summaries, not raw HTML.

URL ingestion

Feed URLs to the extract endpoint and get clean markdown back. Pipe into mem.sh for RAG, or directly into your prompt.

Grounding LLM responses

Search before answering. Retrieve sources before claiming. search.sh is the grounding layer for any agent that needs to verify facts.

API reference

POST  /v1/search       # Web search → structured results + snippets
POST  /v1/search/news  # News search → headlines + summaries
POST  /v1/extract      # URL → clean markdown

Quick start

1POST /v1/search with {"query": "...", "max_results": 5} → web search results
2POST /v1/search/news with {"query": "..."} → recent news articles
3POST /v1/extract with {"urls": ["https://..."]} → clean markdown content

Tips

Set include_answer: true to get an AI-generated summary alongside search results.
Use exclude_domains to remove noise sources (e.g. reddit.com, quora.com).
POST /v1/extract accepts a single URL string or an array of URLs.
Failed extractions (paywalls, 404s) are reported in the failed array — the request still succeeds.
Check Retry-After header on 429 responses before retrying.

Action	Cost	Notes
Web search	$0.005	Tavily ($5/1K calls)
News search	$0.005	Tavily ($5/1K calls)
URL extract	$0.005	Tavily extract (at cost)