Vectara MCP

What is Vectara MCP about?

Vectara MCP exposes two primary tools – ask_vectara and search_vectara – that let AI agents perform RAG queries or plain semantic searches against Vectara indexes. The server implements the Model Context Protocol (MCP), so any MCP‑compatible client (e.g., Claude Desktop) can call these tools without additional glue code.

How to use Vectara MCP?

1. Installation: pip install vectara-mcp.
2. Run the server (as an MCP server) – the package registers its tools automatically when invoked via the MCP runtime (e.g., uv tool run vectara-mcp).
3. Configure the client – add the following to claude_desktop_config.json:

{
  "mcpServers": {
    "Vectara": {
      "command": "uv",
      "args": ["tool", "run", "vectara-mcp"]
    }
  }
}

4. Restart the client application. The hammer icon will reveal the two Vectara tools.
5. When a tool is first invoked, the client prompts for the Vectara API key and one or more corpus keys.

Key features of Vectara MCP

• ask_vectara: RAG query that returns a generated answer plus the underlying search results.
• search_vectara: Pure semantic search returning matching passages without generation.
• Configurable parameters such as number of context sentences, lexical interpolation, result limits, and generation preset selection.
• Works with any MCP‑compatible client, not limited to Claude Desktop.
• Open‑source and distributed via PyPI.

Use cases of Vectara MCP

• Enabling autonomous agents to retrieve up‑to‑date factual information from Vectara‑hosted knowledge bases.
• Providing chat‑based assistants with low‑hallucination RAG capabilities for internal documentation, support tickets, or product FAQs.
• Building custom search‑oriented plugins for AI chat interfaces that require source citations.
• Integrating Vectara search into workflow automation tools that rely on MCP for tool orchestration.

FAQ from Vectara MCP

Q: Do I need a Vectara account? A: Yes, you must have a Vectara API key and at least one corpus key to use the tools.

Q: Can I run multiple corpora simultaneously? A: Provide a list of corpus keys in the corpus_keys argument; the server will search across all supplied corpora.

Q: What language models are used for generation? A: The default generation preset (vectara-summary-table-md-query-ext-jan-2025-gpt-4o) leverages a Vectara‑hosted LLM; you can specify a different preset via the generation_preset_name argument.

Q: How do I change the number of context sentences? A: Use the optional n_sentences_before and n_sentences_after parameters when calling the tool.

Q: Is there a limit on the number of returned search results? A: The max_used_search_results argument caps the number of results used for generation (default 10).

Vectara MCP Server

🔌 Compatible with Claude Desktop, and any other MCP Client!

Vectara MCP is also compatible with any MCP client

The Model Context Protocol (MCP) is an open standard that enables AI systems to interact seamlessly with various data sources and tools, facilitating secure, two-way connections.

Vectara-MCP provides any agentic application with access to fast, reliable RAG with reduced hallucination, powered by Vectara's Trusted RAG platform, through the MCP protocol.

Installation

You can install the package directly from PyPI:

pip install vectara-mcp

Available Tools

• ask_vectara: Run a RAG query using Vectara, returning search results with a generated response.

  Args:

  • query: str, The user query to run - required.
  • corpus_keys: list[str], List of Vectara corpus keys to use for the search - required. Please ask the user to provide one or more corpus keys.
  • api_key: str, The Vectara API key - required.
  • n_sentences_before: int, Number of sentences before the answer to include in the context - optional, default is 2.
  • n_sentences_after: int, Number of sentences after the answer to include in the context - optional, default is 2.
  • lexical_interpolation: float, The amount of lexical interpolation to use - optional, default is 0.005.
  • max_used_search_results: int, The maximum number of search results to use - optional, default is 10.
  • generation_preset_name: str, The name of the generation preset to use - optional, default is "vectara-summary-table-md-query-ext-jan-2025-gpt-4o".
  • response_language: str, The language of the response - optional, default is "eng".

  Returns:

  • The response from Vectara, including the generated answer and the search results.

• search_vectara: Run a semantic search query using Vectara, without generation.

  Args:

  • query: str, The user query to run - required.
  • corpus_keys: list[str], List of Vectara corpus keys to use for the search - required. Please ask the user to provide one or more corpus keys.
  • api_key: str, The Vectara API key - required.
  • n_sentences_before: int, Number of sentences before the answer to include in the context - optional, default is 2.
  • n_sentences_after: int, Number of sentences after the answer to include in the context - optional, default is 2.
  • lexical_interpolation: float, The amount of lexical interpolation to use - optional, default is 0.005.

  Returns:

  • The response from Vectara, including the matching search results.

Configuration with Claude Desktop

Add to your claude_desktop_config.json:

{
  "mcpServers": {
    "Vectara": {
      "command": "uv",
      "args": [
        "tool",
        "run",
        "vectara-mcp"
      ]
    }
  }
}

Usage in Claude Desktop App

Once the installation is complete, and the Claude desktop app is configured, you must completely close and re-open the Claude desktop app to see the Vectara-mcp server. You should see a hammer icon in the bottom left of the app, indicating available MCP tools, you can click on the hammer icon to see more detial on the Vectara-search and Vectara-extract tools.

Now claude will have complete access to the Vectara-mcp server, including the ask-vectara and search-vectara tools. When you issue the tools for the first time, Claude will ask you for your Vectara api key and corpus key (or keys if you want to use multiple corpora). After you set those, you will be ready to go. Here are some examples you can try (with the Vectara corpus that includes information from our website:

Vectara RAG Examples

1. Querying Vectara corpus:

ask-vectara Who is Amr Awadallah?

2. Searching Vectara corpus:

search-vectara events in NYC?

Acknowledgments ✨

• Model Context Protocol for the MCP specification
• Anthropic for Claude Desktop