GPT-5.5 x402 Gateway MCP Server

Remote Streamable HTTP MCP gateway for GPT-5.5 chat plus 92 text, web, JSON, and x402 discovery tools with Base USDC pay-per-call.

Documentation

GPT-5.5 x402 Gateway

Remote MCP and OpenAI-compatible GPT-5.5 gateway with x402 USDC settlement on Base.

The service is built for agent buyers that need low-cost GPT-5.5 calls, deterministic utility tools, and ready-to-use x402/MCP/Bazaar integration resources.

Public Endpoints

  • Service: https://gpt55.558686.xyz
  • MCP Streamable HTTP: https://gpt55.558686.xyz/mcp
  • MCP SSE discovery: https://gpt55.558686.xyz/mcp/sse
  • MCP config: https://gpt55.558686.xyz/mcp/config
  • MCP server metadata: https://gpt55.558686.xyz/server.json
  • x402 discovery: https://gpt55.558686.xyz/.well-known/x402
  • OpenAPI: https://gpt55.558686.xyz/openapi.json
  • Pricing: https://gpt55.558686.xyz/pricing.json
  • Buyer guide: https://gpt55.558686.xyz/buyer-guide.json

Market Positioning

  • Compact GPT-5.5 entry request: $0.0001, or 10,000 compact requests per $1.
  • Long GPT-5.5 request: $0.002134, up to 4096 output tokens.
  • True max-output GPT-5.5 request: $0.066667, up to 128000 output tokens.
  • Utility tools: from $0.0001.
  • x402/MCP/Bazaar integration kits: paid deterministic endpoints for buyers and sellers that need copy-paste-ready integration payloads.
  • Public x402 API listings commonly use per-call pricing around $0.001-$0.01; lead with the $0.0001 compact endpoint for trials, then route larger buyers to /v1/chat/completions/long and /v1/chat/completions/max.
  • Gross revenue per 1M output tokens is about $0.521 across the proportional chat tiers.
  • Upto-capable x402 clients can use token-metered settlement based on actual GPT-5.5 response usage, with a $0.0001 minimum and the selected tier price as the cap.
  • Token-metered target prices at 70%+ gross margin: input about $0.0557 / 1M tokens, cached input about $0.0056 / 1M tokens, output about $0.334 / 1M tokens.

Live quotes from https://gpt55.558686.xyz/pricing.json and unauthenticated HTTP 402 responses are authoritative. Directory pages can lag behind live x402 quote metadata.

Tool Surface

The service exposes 92 MCP tools that route to direct x402-paid HTTP endpoints. Tool calls through the MCP facade return the direct endpoint, pricing, x402 discovery links, and example payload. Paid execution happens on the HTTP endpoint with x402 Base USDC or a private Bearer key; the MCP facade does not bypass payment.

Core endpoints include:

  • POST /v1/chat/completions
  • POST /v1/chat/completions/long
  • POST /v1/chat/completions/max
  • POST /v1/tools/summarize
  • POST /v1/tools/translate
  • POST /v1/tools/rewrite
  • POST /v1/tools/answer
  • POST /v1/tools/code-review
  • POST /v1/tools/extract-json
  • POST /v1/tools/x402-mcp-integration-kit
  • POST /v1/tools/x402-agentcore-integration-kit
  • deterministic utility endpoints for URL metadata, URL text, links, HTTP status, JSON validation, hashing, Base64, timestamps, regex, UUID, and text stats.

Buyer Smoke Tests

MCP initialize:

curl -s https://gpt55.558686.xyz/mcp \
  -H 'content-type: application/json' \
  -d '{"jsonrpc":"2.0","id":1,"method":"initialize","params":{"protocolVersion":"2025-06-18","capabilities":{},"clientInfo":{"name":"smoke","version":"1.0.0"}}}'

x402 quote:

curl -i https://gpt55.558686.xyz/v1/chat/completions \
  -H 'content-type: application/json' \
  -d '{"model":"gpt-5.5","messages":[{"role":"user","content":"Say hi"}],"max_tokens":8}'

Expected unauthenticated response: HTTP 402 with a payment-required header containing Base USDC x402 payment requirements.

Glama / Docker Smoke

This repository includes a small no-secret Docker wrapper for directory checks such as Glama. It starts a local Streamable HTTP MCP endpoint at /mcp, proxies to the public remote MCP endpoint when available, and falls back to local introspection responses for initialize, tools/list, resources/list, and prompts/list.

The wrapper does not handle private keys, wallet signatures, KYC, JWTs, custody, or paid execution. Paid calls still happen through the public x402 HTTP endpoints advertised by the live gateway.

Run locally:

npm start

In another shell:

npm run smoke

Build and run with Docker:

docker build -t gpt55-x402-gateway .
docker run --rm -p 3000:3000 gpt55-x402-gateway

Then submit the repository to Glama using the Dockerfile in this repo. After Glama creates a public server page and score badge, update directory PRs with the generated badge URL.

Local Commands

cd D:\gpt55-token-gateway
npm start
npm run smoke
npm run stats
npm run funnel
npm run monitor

Start a temporary HTTPS tunnel:

D:\gpt55-token-gateway\start-tunnel.ps1

Use the https://...trycloudflare.com URL as the public base URL.

Poe API Bot Settings

  • base_url: tunnel URL, no trailing slash
  • api_type: chat_completions_api
  • model: gpt-5.5
  • api_key: value of PUBLIC_API_KEY in .env
  • pricing.prompt: 0.000005
  • pricing.completion: 0.00003
  • pricing.input_cache_reads: 0.0000005
  • context_pricing: optional; keep the same prices initially until real demand appears.

The gateway only forwards gpt-5.5 upstream, even if a caller passes another model name.

Create Poe Bot

  1. Open https://poe.com/api/keys in a browser where you are logged in.
  2. Create/copy a Poe API key.
  3. Put it in .env as POE_API_KEY=....
  4. Run:
cd D:\gpt55-token-gateway
npm run poe:create

x402 / USDC Settlement

Set these in .env to require on-chain payment for callers that do not use PUBLIC_API_KEY:

X402_ENABLED=true
X402_PAY_TO=0xYourBaseUsdcWallet
X402_NETWORK=eip155:8453
X402_PRICE=$0.0001
X402_LONG_PRICE=$0.002134
X402_MAX_PRICE=$0.066667
X402_TOOL_PRICE=$0.0001
X402_PROBE_PRICE=$0.0001
X402_ENABLE_UPTO=true
X402_UPTO_MARGIN=3.34
X402_UPTO_MIN_USD=0.0001
MAX_PAID_OUTPUT_TOKENS=192
MAX_LONG_PAID_OUTPUT_TOKENS=4096
MAX_TRUE_PAID_OUTPUT_TOKENS=128000

Restart the gateway after changing .env.

Calls with your PUBLIC_API_KEY still bypass the paywall, so Poe/private tests keep working.

Current x402 settings:

network: Base mainnet (eip155:8453)
asset: USDC
pay_to: 0x1f0130669ca6fd02e025a984cc038f139df19a2f
compact_chat_price: $0.0001 per compact chat completion request, max 192 output tokens
long_chat_price: $0.002134 per long chat completion request, max 4096 output tokens
max_chat_price: $0.066667 per max-output chat completion request, max 128000 output tokens
tool_price: $0.0001 per tool request
probe_price: $0.0001 per payment probe
facilitator: CDP x402
schemes: exact, upto
upto_margin: 3.34
upto_minimum: $0.0001

The lowest-cost x402 proof endpoint is available as both GET /v1/x402-ping and POST /v1/x402-ping. It does not call the upstream model.

For callers that support the upto payment scheme, chat endpoints settle by actual token usage after the upstream response reports usage, with the selected fixed tier price as the maximum charge. Callers that only support exact continue to pay the fixed tier price.

Check the payment requirement:

cd D:\gpt55-token-gateway
npm run x402:check

Dry-run the cheapest payment flow without sending funds:

cd D:\gpt55-token-gateway
npm run pay:smoke

pay:smoke defaults to X402_PAY_SMOKE_TARGET=ping-get. Other targets are ping-post, summarize, and chat. It only sends a real payment when both X402_PAYER_PRIVATE_KEY is set and CONFIRM_X402_PAYMENT=true.

Monitor the full discovery and settlement funnel:

cd D:\gpt55-token-gateway
npm run monitor

The monitor writes logs\monitor.jsonl and checks:

  • public health and discovery endpoints
  • x402 402 quote shape and Bazaar extension
  • OpenX402 facilitator whitelist status
  • OpenX402 and CDP Bazaar discovery indexing
  • Base USDC receiver balance
  • local quote/settle/completion funnel

Submit to x402-list for manual review after setting an operator email:

cd D:\gpt55-token-gateway
$env:X402_LIST_CONTACT_EMAIL="[email protected]"
npm run x402list:submit

Submit the public discovery URLs to IndexNow-capable search engines:

cd D:\gpt55-token-gateway
npm run indexnow:submit

The gateway serves the IndexNow key at /<key>.txt. Set INDEXNOW_KEY only if you need a custom stable key; otherwise a deterministic public key is generated from the site URL and pay-to address.

OpenX402 discovery is not only a website crawl. The pay-to address must be accepted by the facilitator, and Bazaar-style discovery normally appears after a successful x402 settlement has carried the resource metadata.

Logs

Usage logs are written to:

D:\gpt55-token-gateway\logs\usage.jsonl

Estimate revenue and cost from the local usage log:

cd D:\gpt55-token-gateway
npm run stats