Plurality AI MCP

Plurality AI MCP Server

Multi-model AI consensus directly in your editor. 17 tools spanning consensus, research, code review, debugging, and more — all powered by 30+ frontier models.

Early access includes all tools and priority onboarding.

What you get

30+ frontier models

Query Claude, GPT, Gemini, Llama, Grok, DeepSeek, Qwen, Mistral, and more in a single call.

Built-in consensus engine

Multi-round deliberation with convergence detection, stance tracking, and synthesis.

Hallucination detection

Cross-model fact-checking flags claims that only one model supports.

Works everywhere

Use in Claude Code, Claude Desktop, Cursor, Windsurf, or any MCP-compatible client.

Expert personas

Domain-specific expert panels for security, medical, legal, finance, engineering, and more.

Web grounding

Claims verified against live web sources with citation links.

17 purpose-built tools

Each tool is optimized for a specific workflow. Use delphi as the unified entry point, or call any tool directly.

Core Deliberation

delphi

Unified entry point — describe what you need and it auto-routes to the right tool

delphi_query = Consensus / Quick Take

Multi-model consensus with convergence detection, hallucination flags, and control comparison

delphi_debate = Debate

Structured debate with sequential or roundtable format and optional adversarial mode

delphi_forecast = Forecast

Multi-model predictions with confidence aggregation and scenario planning

delphi_factcheck

Verify claims against multiple AI perspectives with severity-coded flags

Analysis & Reasoning

delphi_think MCP only

Deep single-model analysis with question enrichment and extended reasoning

delphi_research MCP only

Multi-stage pipeline: research model + web search + synthesis model

delphi_challenge MCP only

Stress-test ideas with multi-model adversarial critique and blind spot detection

delphi_compare MCP only

Side-by-side evaluation of options with independent voting and criteria breakdown

delphi_architect MCP only

System design decisions analyzed across scalability, security, cost, and reliability

delphi_explain MCP only

Multi-model explanations at ELI5, beginner, intermediate, or expert level

delphi_brainstorm MCP only

Generate diverse ideas with automatic deduplication and consensus ranking

Developer Tools

delphi_codereview MCP only

Multi-model code review for bugs, security, performance, and quality

delphi_debug MCP only

Multi-model root cause analysis with confidence-ranked diagnoses

delphi_testgen MCP only

Generate test cases with edge case coverage from multiple QA perspectives

Utilities

delphi_estimate_cost

Preview token costs before running a query

delphi_list_models

List available models in each tier with pricing

Coming Soon

delphi_redteam = Red Team

Adversarial red team analysis — find logic flaws, missing evidence, and implementation risks

delphi_peerreview = Peer Review

Author-reviewer workflow with scoring rubric and optional revision round

delphi_report = Reports

Generate narrative, brief, or memo reports from completed deliberations

claude_desktop_config.json

{
  "mcpServers": {
    "delphi-mcp": {
      "command": "npx",
      "args": ["delphi-mcp"],
      "env": {
        "OPENROUTER_API_KEY": "your-key-here"
      }
    }
  }
}

One-line install. Works with Claude Code, Claude Desktop, Cursor, Windsurf, and any MCP-compatible client.