Ensemble Mcp

What is Ensemble Mcp?

Ensemble Mcp is a paid MCP server that sells a multi-model ensemble capability to AI agents. It runs a three-round pipeline where Claude, GPT, and Gemini independently draft answers, cross-review each other’s drafts, and a master model produces a final attributed answer. Agents pay per request in USDC via the x402 protocol on Base.

How to use Ensemble Mcp?

Deploy the server via Render (one‑click), Fly.io, or locally. Set PAY_TO_ADDRESS and at least one model API key in environment variables. Clients send a POST to /mcp and receive an HTTP 402 Payment Required with a price quote; they must retry with a valid X-PAYMENT header to execute the query.

Key features of Ensemble Mcp

• Three‑round ensemble pipeline (draft → cross-review → master)
• Supports Claude, GPT, and Gemini models
• Pay‑per‑request in USDC via x402 on Base
• One‑click deploy on Render or Fly.io
• Configurable pricing, models, and token limits
• Reports saved per round and model for auditability

Use cases of Ensemble Mcp

• AI agents that need high‑quality, multi‑model synthesized answers
• Applications requiring cross‑validated reasoning from different LLMs
• Paid, usage‑based access to an ensemble without API‑key billing
• Testing and debugging multi‑model pipelines with granular reports

FAQ from Ensemble Mcp

What models are used in the ensemble?

The ensemble uses Claude (via Anthropic), GPT (via OpenAI), and Gemini (via Google) for the draft and cross‑review rounds. A master model merges the final answer.

How does payment work?

Payment is handled by the x402 protocol. The server returns an HTTP 402 Payment Required with a price quote. The client includes a valid X-PAYMENT header, and the facilitator verifies and settles the USDC transfer on Base.

How much does it cost per call?

The default price is $0.05 per call. Underlying model token costs are approximately $0.01–$0.03 per ensemble run at 2k‑token caps. Pricing can be tuned via PRICE_USD and max_tokens.

Can I deploy Ensemble Mcp myself?

Yes. Deploy options include Render (easiest, one‑click), Fly.io (~$5/month), or any Node server. Local development requires Node, a .env file with PAY_TO_ADDRESS and at least one model API key.

Is there a testnet for testing?

Yes. Set X402_NETWORK=base-sepolia and use the default facilitator. Obtain free testnet USDC from the Circle faucet. Use an x402‑capable client (e.g., x402-fetch) pointed at your server’s /mcp endpoint.