openrouter-deep-research-mcp

Overview

OpenRouter Agents MCP Server is a production-ready platform that orchestrates a network of asynchronous AI agents to conduct consensus-backed research. It features an indexed PGlite database built in WebAssembly, hybrid search capabilities, SQL execution, and sophisticated agent communication protocols. The server supports both HTTP and STDIO transport modes and includes advanced features like embedding-based model routing, session management, and knowledge graph exploration.

Try asking AI

After installing, here are 5 things you can ask your AI assistant:

• research

  Async research that returns a job ID

  Try:Research the impact of AI on climate changeTry:Conduct research on quantum computing advancements
• conduct_research

  Synchronous research with streaming results

  Try:Research latest developments in renewable energyTry:Analyze market trends in electric vehicles
• batch_research

  Parallel batch queries for multiple research topics

  Try:Research AI, blockchain, and IoT simultaneouslyTry:Run parallel research on three different medical conditions
• search

  Hybrid BM25+vector search across the knowledge base

  Try:Search for information about machine learningTry:Find articles about sustainable agriculture
• retrieve

  Retrieve data from index or perform SQL query

  Try:Retrieve all documents about climate changeTry:Get recent research reports
• query

  SQL SELECT query with parameters

  Try:Query all reports published after January 2023Try:Find research with confidence score above 0.8
• graph_traverse

  Explore the knowledge graph connections

  Try:Traverse connections from AI researchTry:Explore relationships between climate studies
• graph_clusters

  Find node clusters in the knowledge graph

  Try:Find clusters related to AI ethicsTry:Identify research communities on renewable energy
• list_rails

  List rails, tunnels, routes, and consensus sessions

  Try:Show all active research tunnelsTry:List available research rails
• ping

  Health check for the server

  Try:Check if the research server is onlineTry:Ping the server
• get_server_status

  Get full server diagnostics and status

  Try:Provide server status reportTry:Show system diagnostics
• list_tools

  List all available tools on the server

  Try:What tools are available?Try:Show me all research capabilities

Comparable tools

Installation

npx @terminals-tech/openrouter-agents --stdio

{
  "mcpServers": {
    "openrouter-agents": {
      "command": "npx",
      "args": ["@terminals-tech/openrouter-agents"],
      "env": {
        "OPENROUTER_API_KEY": "sk-or-..."
      }
    }
  }
}

FAQ

What models are supported?

Supports both high-cost models (Claude, GPT-5, Gemini 3) and low-cost models (Gemini Flash, Claude Haiku) with embedding-based routing for optimal selection.

How does the multi-agent research work?

The system decomposes complex queries into sub-queries, assigns them to specialized agents, and synthesizes results through consensus-based protocols with proper citation.

Compare openrouter-deep-research-mcp with