What it does
Turns web research into compact, source-grounded evidence your agent can actually use.
TinySearchLive
Spend tokens on answers, not webpages.
TinySearch searches, crawls, and reranks the web locally, then gives your agent only the source-grounded evidence worth putting in its context.
01 — Overview
Web research that respects the context window.
How it works
Search → page ranking → crawl → chunk ranking → compact evidence.
Use cases
Give agents live web research, cut context waste, and keep every passage tied to a URL.
02 — Install
One command, or the full stack.
Start with a single uvx MCP command and no search service, or deploy the Docker and SearXNG stack for a shared network endpoint.
Fast path
Start the local research layer.
Let any stdio MCP client launch the published package with DDGS and local embeddings.
uvx --from "tinysuite-search[server]" tinysearch03 — Tools
Three MCP tools and a Python API.
The MCP tools return grounded prompts. The Python API returns structured schema-v1 evidence for applications that want to inspect, store, or render the result themselves.
Web research
research(query)
Discover relevant URLs, crawl the best pages, and return ranked evidence chunks.
Known URLs
scrape_url(url, query)
Inspect a known URL and extract query-relevant snippets from that page.
Fresh context
get_current_datetime()
Orient time-sensitive research before calling the web research tools.
Pipeline control
Configuration
Start with DDGS, add SearXNG when needed, and tune embeddings, chunking, and quotas.
04 — Example output
What the agent actually receives.
QUESTION
What should I know about the project?
RESULTS
1. TinySearch
https://github.com/MarcellM01/TinySearch
Self-hosted web research for MCP agents...
INSTRUCTIONS
Answer only from the results. Cite source URLs.Get started
Give your agent evidence, not webpages.
Install TinySearch over stdio in one command, or read the full configuration guide.