mcpField
E
Ensemble Mcp

Ensemble Mcp

Local
github.com/vkrishnavAI & Agents0Updated 19w ago

About Ensemble Mcp

Ensemble Mcp is a paid MCP server that gives AI agents a synthesized answer produced through a three-round exchange among Claude, GPT, and Gemini. Each model drafts, reviews the others' work, and a master model combines the revised reports; payment is per request in USDC via x402 on Base.

Key features

  1. 01Three-round multi-model answer pipeline
  2. 02Independent drafts from Claude, GPT, and Gemini
  3. 03Cross-review and revision between models
  4. 04Master model synthesizes attributed reports
  5. 05Per-request USDC payment via x402

Use cases

  1. 01Request a synthesized answer from multiple models
  2. 02Compare model drafts and cross-reviews
  3. 03Access ensemble queries with per-request payment

MCP support

It exposes the paid `ensemble_query(question)` tool over MCP. Requests are paid in USDC via x402, and the server returns a synthesized result.

Configuration

Local server config is the same for every client. Copy it into your client’s MCP config file.

No install method listed yet

We haven’t collected a package or remote endpoint for this server. Follow the instructions in the project README.

Repository

Tools

No tool list yet

Local servers expose tools only once running on your machine, and the maintainer has not provided a tool list.

Details

From the project README

ensemble-mcp

A paid MCP server that sells one capability to AI agents: a multi-model ensemble. Agents pay per request in USDC via x402 on Base.

What the paid tool does

ensemble_query(question) runs a 3-round pipeline where models learn from each other:

RoundWhat happensReport file
1 — DraftEach model (Claude, GPT, Gemini) answers independentlyreports/<job>/<model>.round1.md
2 — Cross-reviewEach model reads the other models' drafts, critiques, and revises its ownreports/<job>/<model>.round2.md
3 — MasterA master model merges the revised reports into one attributed answerreports/<job>/master.md

Every model gets its own report file per round, and the whole exchange happens over the Model Context Protocol on the outside — pay once, get the synthesized result.

Deploy in one click

Option A — Render.com (easiest)
  1. Push this repo to GitHub.
  2. Click → Deploy to Render
  3. Fill in PAY_TO_ADDRESS and at least one model API key in the Render dashboard.
  4. Done — Render auto-deploys on every push to main.
Option B — Fly.io (~$5/mo, zero idle cost)
brew install flyctl && fly auth login
fly launch --no-deploy          # creates app, detects fly.toml
fly secrets set \
  PAY_TO_ADDRESS=0xYOUR_WALLET \
  ANTHROPIC_API_KEY=sk-ant-... \
  OPENAI_API_KEY=sk-...         # optional
fly deploy                      # builds + deploys in ~90 s
Option C — local dev
cp .env.example .env   # fill in PAY_TO_ADDRESS + at least one model API key
npm install
npm run dev
  • GET / and GET /health are free (service discovery).
  • POST /mcp returns HTTP 402 Payment Required with a price quote until the client retries with a valid X-PAYMENT header. The facilitator verifies and settles the USDC transfer, then the request goes through.

Testing the payment flow (testnet, free)

  1. Keep X402_NETWORK=base-sepolia and the default facilitator.
  2. Get free testnet USDC from the Circle faucet (Base Sepolia).
  3. Use an x402-capable client — e.g. x402-fetch or x402-axios from the x402 repo — pointed at http://localhost:3000/mcp, or wire it into an MCP client config.

Going live (real revenue)

  1. Set X402_NETWORK=base and switch FACILITATOR_URL to Coinbase CDP's mainnet facilitator (see docs.cdp.coinbase.com/x402).
  2. Deploy anywhere Node runs — Render, Fly.io, Railway, a VPS. (Vercel/Cloudflare also support x402 natively; this codebase is plain Express so a small adapter is needed for Workers.)
  3. List the server in x402/MCP discovery indexes (e.g. the x402 Bazaar) so agents can find it.

USDC lands directly in PAY_TO_ADDRESS — no invoicing, no Stripe, no API-key billing.

Pricing math

At $0.05/call, your cost is the underlying model tokens (~$0.01–0.03 per ensemble run at 2k-token caps). Tune PRICE_USD, max_tokens, and the model list for margin.

Extending

  • Add members: edit getEnsembleMembers() in src/providers.ts.
  • Add more paid tools: register them in buildMcpServer() in src/index.ts and add routes to the paymentMiddleware config (different prices per tool/route are supported).
  • Persist reports to S3/R2 instead of local disk: swap saveReport() in src/ensemble.ts.

Comments

Markdown supported

New to MCP?

From what the protocol is to installing your first server, these answers cover most first questions.

An open protocol that defines how AI applications connect to external tools and data in a uniform way. Clients talk to the model; MCP servers provide tools, resources and prompt templates. They exchange standard messages, so one server can be reused by many clients.