Installation

Installation Guide

Setup VerifiMind™ PEAS for Claude, Cursor, VS Code, and other MCP clients

Last Updated: May 17, 2026 — v0.5.34, Phase 90 "Adoption First"

All 13 tools free forever under the Core Tools Always Free pledge. No registration required. No payment ever.

---

Prerequisites

• An MCP-compatible client (Claude Desktop, Claude Code, Cursor, VS Code, Windsurf, ChatGPT Codex CLI, OpenAI Agents SDK, etc.)
• Node.js 18+ (for npx mcp-remote bridge — only if your client requires stdio transport)

---

Critical Configuration Requirements

Before you copy any example below, three things to know:

1. URL must end with trailing slash: /mcp/ (not /mcp)
   • Without the trailing slash, you get a 308 Permanent Redirect to /mcp/
   • Programmatic SDK clients (python-httpx, node, requests, axios) often fail to follow POST redirects per RFC 7538 → silent connection failure
2. Transport must be streamable-http (NOT http-sse, NOT sse)
   • This is the current MCP 2025-03-26 spec
   • Legacy SSE is not supported
3. POST requests must include both Accept types: Accept: application/json, text/event-stream
   • Missing the event-stream side returns 406 Not Acceptable
   • Managed clients (Claude/Cursor/VS Code/Agents SDK) handle this automatically; custom clients must set both

---

Connection Methods

Method 1 — Streamable-HTTP (recommended for native clients)

For clients that support streamable-http transport natively (Cursor, VS Code, Windsurf, Codex CLI).

URL: https://verifimind.ysenseai.org/mcp/

Cursor / VS Code / Windsurf

Add to .cursor/mcp.json or .vscode/mcp.json:

{
  "servers": {
    "verifimind": {
      "url": "https://verifimind.ysenseai.org/mcp/",
      "transport": "streamable-http"
    }
  }
}

---

Method 2 — mcp-remote Bridge (for stdio-only clients)

For clients that require stdio transport (Claude Desktop, ChatGPT Codex CLI workarounds).

Claude Code (CLI) — one command

claude mcp add -s user verifimind -- npx -y mcp-remote https://verifimind.ysenseai.org/mcp/

After this, restart Claude Code and type /mcp to see the 13 available tools.

Claude Desktop

Add to your config file:

OS	Config Path
macOS	~/Library/Application Support/Claude/claude_desktop_config.json
Windows	%APPDATA%\Claude\claude_desktop_config.json
Linux	~/.config/Claude/claude_desktop_config.json

{
  "mcpServers": {
    "verifimind": {
      "command": "npx",
      "args": ["-y", "mcp-remote", "https://verifimind.ysenseai.org/mcp/"]
    }
  }
}

Restart Claude Desktop after saving.

---

Method 3 — Direct HTTP (custom MCP clients)

For any custom MCP client using direct HTTP:

{
  "mcpServers": {
    "verifimind-genesis": {
      "url": "https://verifimind.ysenseai.org/mcp/",
      "transport": "streamable-http"
    }
  }
}

For Python developers using the OpenAI Agents SDK:

import asyncio
from agents import Agent, Runner
from agents.mcp import MCPServerStreamableHttp

async def main():
    async with MCPServerStreamableHttp(
        name="VerifiMind Genesis",
        params={"url": "https://verifimind.ysenseai.org/mcp/"},
        cache_tools_list=True,
    ) as server:
        agent = Agent(
            name="Validator",
            instructions="Use VerifiMind tools to validate concepts.",
            mcp_servers=[server],
        )
        result = await Runner.run(agent, "Use run_full_trinity to validate the concept 'AI-powered code review'")
        print(result.final_output)

asyncio.run(main())

pip install openai-agents required.

---

Verification

After connecting, verify the server is working:

1. Health Check

curl https://verifimind.ysenseai.org/health

Should return JSON with "status": "healthy", "version": "0.5.34", "inference_mode": "live", and "tools": 13.

2. Tools List

Your MCP client should show 13 tools:

• Trinity validation (4) — consult_agent_x, consult_agent_z, consult_agent_cs, run_full_trinity
• Template library (6) — list_prompt_templates, get_prompt_template, export_prompt_template, register_custom_template, import_template_from_url, get_template_statistics
• Coordination tools (3, new in v0.5.16) — coordination_handoff_create, coordination_handoff_read, coordination_team_status

3. Test Query

Try in your MCP client:

"Use consult_agent_x to briefly analyze the AI validation market for MCP-based tools in 2026"

Or test the complete Trinity pipeline:

"Use run_full_trinity to validate the concept of an AI-powered code review tool"

---

Troubleshooting

Common Issues — Quick Reference

Issue	Likely Cause	Solution
308 / silent connection fail	Missing trailing slash on /mcp	Use /mcp/ (with trailing slash)
307 / wrong landing page	Client pointed at / not /mcp/	Configure full path including /mcp/
406 Not Acceptable	Missing dual Accept header	Send Accept: application/json, text/event-stream
Tools not showing	Client cached old schema	Remove and re-add the MCP server connection
403 Forbidden	Network-level OR Sentinel blocklist	Try different network; if still blocked, file Issue
Connection timeout	Firewall / proxy	Check HTTPS access to verifimind.ysenseai.org
"Session expired"	Ephemeral session by design	Reconnect — sessions are stateless
202 Accepted	NOT an error — normal streamable-http response	No action needed; client should continue listening on SSE

Full Troubleshooting Guide

See docs/MCP_Server_Troubleshooting_Guide.md for detailed per-status-code diagnostics, client-specific notes (Codex CLI bug workaround, Smithery auth issue, etc.), and transport-type clarifications.

---

Official MCP Registry

VerifiMind PEAS is listed on the official MCP Registry: registry.modelcontextprotocol.io/?q=verifimind

Current published version: v3.11.0 (server.json), aligned with server v0.5.34.

---

What you can do once connected

The 13 tools let you:

• Validate any concept through the X → Z → CS Trinity (innovation + ethics + security in one call)
• Run per-agent BYOK — mix Gemini + Claude + GPT across the three agents in one Trinity call
• Browse and apply the prompt template library (X / Z / CS agent templates, taggable, versioned)
• Use the MACP coordination layer to create / read handoffs between AI agents in multi-agent FLYWHEEL workflows
• Read live methodology context from Resources (Genesis Master Prompt v4.2, validation history, project metadata)

Full features documentation: MCP_SERVER_FEATURES.md

---

Phase 90 Context

This server is in Phase 90 "Adoption First" — monetization is on STANDBY. The pricing surface that earlier versions of this guide referenced (Pioneer $9/month tier) is no longer active. All 13 tools are free, and the Beta v0.6.0 milestone has been redefined as the Evaluation Roadmap M0 + M1 (credibility milestone, NOT commercial milestone).

If you want to help build the labeled evaluation dataset for the methodology (Evaluation Roadmap M2), register as an Early Adopter — UUID-only, consent-first, opt out anytime.

---

← Back to Home