Agent♥︎Age
Catalog

Grounder

Official

by rozetyp · Python

Web grounding for local & cloud LLMs: search, fetch, token-capped evidence packs, and research.

io.github.rozetyp/grounder (Grounder MCP)

Grounder MCP provides “live web grounding” for local and cloud LLMs via four MCP tools. It supplies ranked web search results, real page content, and a cited evidence pack that is token-capped to fit the model’s context window.

🛠️ Key Features

  • Web grounding for local & cloud LLMs
  • Four MCP tools for search, fetch, and evidence pack retrieval
  • Ranked results and real page content
  • Cited evidence packs sized to the context window
  • Implemented as a thin stdio client for the hosted service at grounder.dev

🚀 Use Cases

  • Retrieval and RAG workflows needing current web context
  • Enriching LLM responses with cited, token-capped evidence
  • Using model-context-protocol with clients like Claude Desktop, Cursor, LM Studio, and Continue.dev

⚡ Developer Benefits

  • Works with MCP clients including Claude Desktop, Cursor, LM Studio, and Continue.dev
  • No heavy local browser/scraper components mentioned; relies on hosted grounding
  • Integrates model-context-protocol (MCP) for grounding pipelines

⚠️ Limitations

  • Hosted grounding is provided because “every model is frozen at its training cutoff,” so Grounder fetches current web content externally

Topics

ai-agentsclaudecursorgroundingllmlm-studiolocal-llmmcpmcp-servermodel-context-protocolollamaragretrievalsearch-apiweb-search
Grounder - agentage MCP Catalog