The search tool scores and classifies what it returns and everything comes back as markdown rather than HTML, which is the difference between an agent reading three good sources and an agent burning its context on navigation chrome. research_topic collapses the whole loop — search, rank, dedupe, extract, synthesise — into one call, and the pricing follows the same logic: one credit per search, extraction free.
A web-research server, published by its author as Nimrod. Four tools: quality-scored Google search, single-page extraction to markdown, up to five pages at once, and a full research pass in one call.
- google_search — organic Google results with quality scores, authority ratings and source-type classification, with date, site and language control
- extract_webpage_content — any URL turned into clean markdown with the boilerplate stripped
- extract_multiple_webpages — up to five URLs fetched concurrently
- research_topic — search, rank, dedupe, extract and synthesise, returned from a single call
- A desktop build that runs extraction on your own machine, bundled with six research skills and a deep-research agent
- Listed in the MCP Registry as ai.orchis/nimrod, so registry-aware clients can install it without a config file
An account with the service; new accounts start with a 50-credit trial and no card. The hosted endpoint is https://nimrod.orchis.ai/mcp — add it as a custom connector and sign in, or pass an Authorization: Bearer key. Desktop builds exist for Windows, macOS and Linux; the macOS build is not Apple-signed, so it needs one xattr -d com.apple.quarantine before it will run. Search costs 1 credit and research_topic costs 1 credit plus one per focus area; extraction is free on every plan.