Skillget

nodebench-ai

mcp
SUMMARY

NodeBench AI is an autonomous research platform. It acts as a team of AI researchers that scour the web, analyze documents, and synthesize decision-ready briefs for founders pitching VCs or investors doing diligence—all fully backed by source evidence and receipts instead of hallucinations.

README.md

NodeBench AI

Operating intelligence for founders. Search any company, get a structured intelligence packet, delegate to your agents.

Live: nodebenchai.com

What it does

  1. Ask anything — Type a company name, paste meeting notes, or ask a strategic question
  2. Get your answer — A structured intelligence packet shaped for your role (founder, investor, banker, CEO, legal, student)
  3. Share and delegate — Copy a shareable link, hand off to Claude Code, or delegate to your team

Every result includes: entity truth, signals, risks, contradictions, next actions, and source evidence.

Quick start

# 1. Clone and install
git clone https://github.com/HomenShum/nodebench-ai.git
cd nodebench-ai
npm install

# 2. Set up environment
cp .env.example .env.local
# Add your GEMINI_API_KEY (required for search)

# 3. Run
npm run dev
# Open http://localhost:5191

MCP server (use in Claude Code, Cursor, Windsurf)

# Claude Code
claude mcp add nodebench -- npx -y nodebench-mcp

# Cursor / Windsurf — add to MCP config:
{ "command": "npx", "args": ["-y", "nodebench-mcp"] }

Architecture

Search / Upload → Entity extraction → Signals + Risks → Artifact Packet → Share / Delegate

7 pages

Page Path Purpose
Ask / Search, upload, get intelligence packets
Memo /deep-sim Decision workbench — variables, scenarios, interventions
Research /research Track what changed, daily brief, signals
Workspace /workspace Documents, notes
Dashboard /founder Company truth, what changed, next moves
Coordination /founder/coordination Peer presence, task delegation, messaging
Entities /founder/entities Competitors, partners, watchlist

Stack

  • Frontend: React + Vite + Tailwind CSS
  • Backend: Convex (real-time database + serverless functions)
  • Search: Gemini (entity extraction) + Linkup (web search)
  • MCP server: 350 tools across 57 domains (packages/mcp-local/)
  • Deployment: Vercel (frontend + API) + Convex (backend)

Key directories

src/                          # React frontend
  features/founder/           # Founder dashboard, coordination, entities
  features/controlPlane/      # Search canvas, result workspace
  features/research/          # Research hub, signals, briefing
packages/mcp-local/           # MCP server (350 tools)
server/                       # Express server (search API, shared context, sync bridge)
convex/                       # Convex backend (schema, operations, workflows)

Environment variables

Only GEMINI_API_KEY is required for the search pipeline. See .env.example for all options.

Variable Required Purpose
GEMINI_API_KEY Yes Entity extraction + grounding
VITE_CONVEX_URL For backend Convex deployment URL
LINKUP_API_KEY Recommended Web search
OPENAI_API_KEY Optional Embedding fallback
ELEVENLABS_API_KEY Optional Voice output

Development

npm run dev              # Start Vite dev server (port 5191)
npm run build            # Production build
npx tsc --noEmit         # Type-check
npx vitest run           # Run tests (from packages/mcp-local/)

License

MIT

Reviews (0)

No results found