# gemy411/multi-research-agents [Health: Active]

**Category:** 🔎 Search & Data Extraction  
**Repository:** https://github.com/gemy411/multi-agents-research  
**GitHub Stars:** 16  
**Views:** 5  
**Installs:** 0  
**Upvotes:** 0  
**Directory Page:** https://allmcps.com/mcp/gemy411-multi-research-agents

## Description
a KTOR server/ MCP server written in Kotlin applying multi-agents schools in a flexible research system to be used with coding or for research any general case.

## Claude Desktop Quick Installation
Heuristic fallback — verify the package name and runner against the repository README before running it. Uses `npx` (confidence: low):

```json
"mcpServers": {
  "multi-research-agents": {
    "command": "npx",
    "args": ["-y","gemy411-multi-research-agents"],
    "env": {
      "OPEN_ROUTER_API_KEY": ""
    }
  }
}
```

**Requires environment variables:** `OPEN_ROUTER_API_KEY` — the values above are empty placeholders; fill in real credentials before running (see the repository for what each one is for).

## Documentation

## What gemy411/multi-research-agents MCP server does

The gemy411/multi-research-agents MCP server provides a coordinated workflow for complex research questions. It exposes a `DeepResearch` MCP tool that accepts a natural-language `query`, and it also provides a REST endpoint at `GET /search?query=...`. Both interfaces invoke the same lead research agent and return the resulting synthesis.

The system is intended for investigations that benefit from dividing a question into separate areas. Examples in the project include comparing quantum-computing developments and examining the economic, environmental, and social effects of urban vertical farming.

## How it works

A lead agent first interprets the incoming question and determines a research plan. It uses a `deployAgent` capability to create specialized subagents, which can work on different aspects of the query in parallel. Each subagent performs focused web searches, retrieves content from selected URLs, and reports its findings to the lead agent.

The lead agent reviews those intermediate results and may deploy more subagents when the investigation needs additional comparisons, timelines, or technical analysis. It then combines the collected findings into a final response. The lead agent uses a Gemini Pro model and supports up to 50 iterations; subagents use Gemini Flash and support up to 30 iterations. Conversation history can be compressed when it exceeds 20 messages.

## Setup and configuration

The project is implemented in Kotlin with Ktor and uses the Koog agent framework. To configure it, copy the example environment file to `.env` and provide an OpenRouter credential:

```bash
cp example.env .env
```

Set `OPEN_ROUTER_API_KEY` in that file. `PORT` can be used to select the server port; the documented default is 8080, while the example configuration uses 4321. Build and start the application from the repository with:

```bash
./gradlew build
./gradlew run
```

The documented startup address is `http://0.0.0.0:8080` when the default port is used.

## Tools and capabilities

The gemy411/multi-research-agents MCP server exposes the following documented research-facing capability:

- `DeepResearch`: accepts a complex research question through MCP and runs the lead-agent workflow.
- Web search and content retrieval: available to research subagents through the Sonar Search Manager.
- Parallel subagent execution: allows separate research tasks to proceed simultaneously.
- Iterative delegation: lets the lead agent add research tasks after reviewing early findings.
- Result synthesis: combines specialized reports into one response.

OpenRouter is used to access the configured language models, including Sonar by Perplexity for search functionality.

## Limitations and notes

The provided material does not document authentication behavior for the MCP transport, supported MCP client applications, deployment instructions, or a project license. Running the application requires an OpenRouter API key, and the available research behavior depends on the configured models and Sonar search access. The REST and MCP interfaces are described as exposing the same underlying functionality; no separate REST-only or MCP-only research features are specified.

_Full upstream README: https://allmcps.com/mcp/gemy411-multi-research-agents/readme_