Metadata-Version: 2.4
Name: grounder-mcp
Version: 0.3.1
Summary: Live web search, page fetch, and a token-capped cited evidence pack as MCP tools for local and cloud LLMs.
Project-URL: Homepage, https://grounder.dev
Keywords: mcp,web-search,fetch,deep-search,rag,grounding,llm,ollama,lm-studio
Requires-Python: >=3.10
Description-Content-Type: text/markdown
Requires-Dist: mcp>=1.2
Requires-Dist: httpx>=0.27

# Grounder

From training data to current knowledge. Gives a local or cloud LLM live web grounding
as four MCP tools: `web_search` (live Google results), `fetch` (a result page as clean,
token-capped markdown), `deep_search` (a cited, token-capped evidence pack that fits a
small context window), and `research` (an agentic search-and-read loop for hard, multi-part
questions). Privacy-first: the hosted service stores no query content.

This package is a thin MCP client; the retrieval and billing run on the hosted service,
so it installs with no browser and runs fine on a laptop.

## Use

```json
{
  "mcpServers": {
    "Grounder": {
      "command": "uvx",
      "args": ["grounder-mcp"],
      "env": { "GROUNDER_API_KEY": "your_key_here" }
    }
  }
}
```

Get a key at https://grounder.dev. Flat monthly plans billed in pages (a search, a fetch, or a
page `deep_search` reads); `research` runs on a small monthly allowance. You pay only for pages
delivered.

<!-- mcp-name: io.github.rozetyp/grounder -->
