Skip to main content
Glama
goofrey

Zoom Search

by goofrey
OverviewSchemaSourceRelated ServersDiscussionsAdmin
Python
Remote

Zoom Search

Zoom Search is a search and evidence tool for AI agents. It helps agents rewrite search questions, gather broader web evidence, zoom into high-value source domains, and return sourced answers with metrics.

It is built for agentic applications that need stronger source discovery, traceability, and answer grounding than a single search call.

Why Zoom Search

  • Agent search tool: expose structured answers, sources, warnings, and metrics for tool-calling agents.

  • Better evidence gathering: rewrite agent questions into stronger search variants.

  • Source-domain zoom-in: search broadly first, then focus on high-value domains.

  • Traceable outputs: preserve source domains, duplicate provenance, warnings, and runtime metrics.

  • MCP/LangGraph ready: use Zoom Search through MCP or LangGraph integrations.

  • Provider-flexible: use built-in engines or custom OpenAI-compatible and native HTTP providers.

Related MCP server: MyWebSearch MCP

Install

pip install zoom-search

Quickstart

Run a deterministic local demo without API keys:

import asyncio

from zoom_search import search


async def main() -> None:
    response = await search(
        question="What hotels in Shenzhen have rooms with exercise bikes?",
        demo_mode=True,
        output_mode="answer_with_sources",
        seed=7,
    )
    print(response.answer)
    print(response.results)


asyncio.run(main())

Agent Tool Example

Install the MCP extra:

pip install "zoom-search[mcp]"

Add Zoom Search to your MCP client:

{
  "mcpServers": {
    "zoom-search": {
      "command": "zoom-search-mcp",
      "env": {
        "ZOOM_SEARCH_LLM_ENGINE": "gemini",
        "ZOOM_SEARCH_LLM_MODEL": "gemini-2.5-flash",
        "ZOOM_SEARCH_LLM_API_KEY": "YOUR_GEMINI_API_KEY",
        "ZOOM_SEARCH_SEARCH_ENGINE": "tavily",
        "ZOOM_SEARCH_SEARCH_API_KEY": "YOUR_TAVILY_API_KEY"
      }
    }
  }
}

Your agent can then call the zoom_search tool with a question argument:

{
  "question": "Which vector databases support hybrid search and metadata filtering for Python apps?",
  "output_mode": "answer_with_sources"
}

The tool returns sourced answers, source-domain zoom-in, warnings, and runtime metrics.

Or wrap it as a LangGraph/LangChain tool:

import os

from langchain.tools import tool

from zoom_search import search


@tool
async def zoom_search_evidence(query: str) -> dict:
    response = await search(
        question=query,
        llm_engine=os.environ["ZOOM_SEARCH_LLM_ENGINE"],
        llm_model=os.environ["ZOOM_SEARCH_LLM_MODEL"],
        llm_api_key=os.environ["ZOOM_SEARCH_LLM_API_KEY"],
        search_engine=os.environ["ZOOM_SEARCH_SEARCH_ENGINE"],
        search_api_key=os.environ["ZOOM_SEARCH_SEARCH_API_KEY"],
        output_mode="answer_with_sources",
    )
    return response.to_dict()

See docs/agent-integration.md for MCP client configuration and provider environment variables.

Benchmarks

Historical evaluations compare direct search against the Zoom Search agent workflow, showing better useful result coverage and stronger final answers with bounded extra time and token cost.

Case

Good results

Answer quality

Extra time

Extra tokens

Playwright authentication reuse

5 -> 7

6.6 -> 8.7

+5.89s

+2,324

GitHub Actions secrets inherit

1 -> 4

2.0 -> 7.8

+8.93s

+2,936

Hydrangea pruning comparison

4 -> 12

7.2 -> 8.4

+12.17s

+5,073

See the full benchmark notes in docs/benchmarks.md.

Runnable examples for demo mode, streaming, conversation history, and LangGraph are available in the examples/ directory.

Documentation

License

Zoom Search is open source under the MIT License.

Available Tools

1 tool

Tool Schema Changelog

Recent tool additions, removals, and schema changes observed during successful MCP inspections.

  1. 1 tool updatev0.1.4
    • First observedzoom_search
Deploy Server
One endpoint for every MCP toolRayrun runs every MCP server behind one URL and holds the credentials. Each call is decided per client, tool, and argument, then recorded.ray.runAd

TDQS

B3.2/5.0

Scored across 1 tool

Disambiguation5/5

With only one tool, there is no risk of ambiguity; the tool's purpose is entirely distinct by default.

Naming Consistency5/5

The single tool uses a clear snake_case naming convention, which is consistent by default.

Tool Count3/5

The server has only one tool, which is on the low end of tool count. While it serves a focused purpose, it feels slightly thin for a server that could potentially offer more related operations.

Completeness4/5

The single tool covers the core search functionality and returns comprehensive results (answer, sources, warnings, metrics, evidence). However, there might be missing features like search configuration or history, but the current scope is reasonable.

Maintenance

ActivityStale
ResponsivenessNo issues

Resources

Related MCP Connectors

Related MCP Servers

  • qsearch

    theYahia
    A
    license
    Not graded
    quality
    B
    maintenance
    Enables AI agents to perform web searches with full content retrieval and multi-engine provenance, including trust scoring and local corpus persistence, via MCP integration.
    3 npm
    2
    Apache 2.0

  • MyWebSearch MCP

    iuiu-py
    F
    license
    Not graded
    quality
    C
    maintenance
    Enables AI agents to perform unified web research through a single MCP server, including search, page fetching, recursive crawling, document parsing, YouTube transcript extraction, and deep multi-query research.
    2
    -

  • jiro

    DevAnimecx
    A
    license
    Not graded
    quality
    A
    maintenance
    Enables AI agents to perform live web searches across 9 engines, scrape web pages into clean formats, and run agentic research with citations via MCP.
    1
    MIT

  • trawlia-mcp

    ayhoung
    A
    license
    Not graded
    quality
    B
    maintenance
    Enables AI agents to search the live web and extract readable page content as MCP tools, with ranked results, domain filters, and news support.
    39 npm
    MIT