惯性聚合 高效追踪和阅读你感兴趣的博客、新闻、科技资讯
阅读原文 在惯性聚合中打开

推荐订阅源

SecWiki News
SecWiki News
H
Help Net Security
罗磊的独立博客
Stack Overflow Blog
Stack Overflow Blog
M
MIT News - Artificial intelligence
Jina AI
Jina AI
L
LangChain Blog
K
Kaspersky official blog
I
Intezer
Martin Fowler
Martin Fowler
爱范儿
爱范儿
AWS News Blog
AWS News Blog
The Hacker News
The Hacker News
Recorded Future
Recorded Future
人人都是产品经理
人人都是产品经理
H
Hackread – Cybersecurity News, Data Breaches, AI and More
C
CXSECURITY Database RSS Feed - CXSecurity.com
Spread Privacy
Spread Privacy
Simon Willison's Weblog
Simon Willison's Weblog
U
Unit 42
N
News and Events Feed by Topic
A
Arctic Wolf
G
GRAHAM CLULEY
Microsoft Azure Blog
Microsoft Azure Blog
博客园 - 聂微东
F
Fortinet All Blogs
C
Cisco Blogs
美团技术团队
Vercel News
Vercel News
K
KPMG report finds enterprise disconnect between AI and its ROI | CIO
H
Hacker News: Front Page
T
Tailwind CSS Blog
I
InfoQ
宝玉的分享
宝玉的分享
Google DeepMind News
Google DeepMind News
博客园 - 司徒正美
P
Palo Alto Networks Blog
A
About on SuperTechFans
Threat Intelligence Blog | Flashpoint
Threat Intelligence Blog | Flashpoint
云风的 BLOG
云风的 BLOG
TaoSecurity Blog
TaoSecurity Blog
Google Online Security Blog
Google Online Security Blog
Exploit-DB.com RSS Feed
Exploit-DB.com RSS Feed
P
Privacy & Cybersecurity Law Blog
H
Heimdal Security Blog
cs.CV updates on arXiv.org
cs.CV updates on arXiv.org
Hacker News: Ask HN
Hacker News: Ask HN
O
OpenAI News
博客园 - Franky
Scott Helme
Scott Helme

DEV Community

Authentication Security Deep Dive: From Brute Force to Salted Hashing (With Java Examples) Why AI Systems Don’t Fail — They Drift Spilling beans for how i learn for exam😁"Reinforcement Learning Cheat Sheet" I Replaced Chrome with Safari for AI Browser Automation. Here's What Broke (and What Finally Worked) How Python Borrows Other People's Work The $40 Architecture: Processing 1 Billion API Requests with 99.99% Uptime Vibe Coding: A Workflow Guide (From Zero to SaaS) Most webhook security guides protect the wrong side. The scary part is delivery. Headless CMS for TanStack Start: Build a Blog with Cosmic EU Age Verification App "Hacked in 2 Minutes" — What Actually Happened Comfy Cloud’s delete function does not actually remove files Running AI Models on GPU Cloud Servers: A Beginner Guide Event-driven media intelligence with AWS Step Functions and Bedrock I scored 500 AI prompts across 8 quality dimensions — here's what broke How to Call Google Gemini API from Next.js (Free Tier, No Backend Needed) The Portal Protocol: Reclaiming Human Connection in the Age of AI How to Fix Your Team's Scattered Knowledge Problem With a Self-Hosted Forum Intro to tc Cloud Functors: A Graph-First Mental Model for the Modern Cloud Designing Multi-Tenant Backends With Both Ownership and Team Access I Built a Neumorphic CSS Library with 77+ Components — Here's What I Learned PostgreSQL Performance Optimization: Why Connection Pooling Is Critical at Scale Cómo construí un SaaS multi-rubro para gestionar expensas en Argentina con FastAPI + Vue 3 🚀 I Built an Ethical Hacking Scanner Tool – Open Source Project I Replaced /usage and /context in Claude Code With a Single Statusline A Pythonic Way to Handle Emails (IMAP/SMTP) with Auto-Discovery and AI-Ready Design I Collected 8.9 Million Polymarket Price Points — Here's What I Found About How Markets Really Move EcoTrack AI — Carbon Footprint Tracker & Dashboard Everyone's Using AI. No One Agrees How. 5 self-hosted ebook managers worth trying in 2026 Building Your First AI Agent with LangChain: From Chatbot to Autonomous Assistant Common SOC 2 Failures (Real World) Stop Vibe-Checking Your AI App: A Practical Guide to Evals How to Use SonarQube and SonarScanner Locally to Level Up Your Code Quality Your Next To-Do App Is Dead — I Replaced Mine with an OpenClaw AI Sign a Nostr event in 60 lines of Python using coincurve — no nostr-sdk, no nbxplorer, no rust toolchain ITGC Audit Explained Like You’re in Big 4 Patch Tuesday abril 2026: Microsoft parcha 163 vulnerabilidades y un zero-day en SharePoint Stop scraping everything: a better way to track competitor price changes Listing on MCPize + the Official MCP Registry while routing payments OUTSIDE the marketplace — how I kept 100% of my x402 revenue Building an AI-Powered Risk Intelligence System Using Serverless Architecture Why We Ripped Function Overloading Out of Our AI Toolchain Testing AI-Generated Code: How to Actually Know If It Works SaaS Churn Is Killing Your Business. Here Is What to Do About It (Without a Support Team) The Speed of AI Is No Longer Linear - And Self-Improving Models Are Why How to Implement RBAC for MCP Tools: A Practical Guide for Engineering Teams From Standard Quote to Persuasive Proposal: AI Automation for Arborists I built a CLI that scaffolds complete multi-tenant SaaS apps Axios CVE-2025–62718: The Silent SSRF Bug That Could Be Hiding in Your Node.js App Right Now The dashboard that ended our friendship Data Pipelines Explained Simply (and How to Build Them with Python) The Hidden Cost of AI Systems Nobody Talks About. undefined vs undeclared, and how typeof behaves Switching from file-based jobs to NATS/Kafka in Rust without changing code io_uring Adventures: Rust Servers That Love Syscalls Why Agentic AI is Killing the Traditional Database The POUR principles of web accessibility for developers and designers Quantum Neural Network 3D — A Deep Dive into Interactive WebGL Visualization How To Install Caveman In Codex On macOS And Windows Automation Pipeline Reliability: Why Your Workflow Breaks When Nobody Is Watching I Built an 'Open World' AI Coding Agent — It Works From ANY Folder From Freelancing to Product: A Tech Service Company's SaaS Transformation China's AI Giants: Adding Tencent Hunyuan & ByteDance Doubao to AI University (74 Providers) On the Vibe Coders and Their Lies clerk: Auto-Summarize Your Claude Code Sessions AI Weekly — 2026/04/10–04/17 | The Model Lockdown Is Here, but the Toolchain Is the Real Battleground AI 週報 — 2026/04/10–2026/04/17 模型封鎖潮來了,但工具鏈才是真戰場 Maybe this is how Open-Source apps are born... 🚀 Fine-Tune LLMs with LoRA and QLoRA: 2026 Guide tRPC v11 + Next.js App Router: End-to-End Type Safety Without the Boilerplate ShadCN UI in 2026: Why I Stopped Installing Component Libraries and Started Owning My Components SaaS Billing in React Server Components: Stripe + Supabase Without a Single `useEffect` Join our DEV Weekend Challenge — $1,000 in Prizes Across TEN winners! Submissions Due April 20 at 6:59 AM UTC. Implementing FSRS Spaced Repetition in Flutter + Supabase — Adding Memory Science to an AI Learning App "I Texted My Localhost From the Train — Claude Code Fixed the Bug Before I Got Home" I Built a Sales Prep AI and It Went Deeper Than Expected Design to Code #2: One JSON, Eleven Outputs Solving the 100M-Row Problem: A Summary Table Pattern for High-Volume Push Notification Logs Flutter Web With Wasm: What Actually Changes For Developers I Built 50 Royalty-Free Soundtracks for My Side Project in a Weekend Using AI Music Generation The Vibe Coding Security Checklist: 7 Things to Check Before You Ship Stop Letting Googlebot Guess Fix Your React App's SEO Right Desconstruindo o Streaming do LinkedIn: Como Criar um Engine de Extração de Vídeo de Alta Performance com HLS e FFmpeg (EDA Part-1) EDA (Exploratory Data Analysis) Explained With Real Life — Why Looking at Your Data Is the Most Important Step in Machine Learning Brand Relationship Management at Scale: Our 4-Touch Outreach System for 200+ Brands Why String.fromEnvironment() Might Return an Empty String in Dart JGuardrails 1.0.0 — Hardening Java LLM Apps Against Jailbreaks, Toxicity, and Prompt Injection Plan and Schedule a Full Week of Threads Content From One Claude Conversation Coding Cat Oran Ep3, Five Tables Changed Everything Updated: BFF Pattern I'm done watching freelancers get buried by 200 proposals. So I'm building the alternative. This is my first post BFS Algorithm in Java Step by Step Tutorial with Examples Tracking LLM Pricing Monthly: An Open Dataset for 22 AI Models How We Measure Content ROI on a Comparison Site: Revenue Attribution Without Perfect Data Introducing Nova AI Ops: The AI-Native Operating System for SRE Teams I built a free desktop video downloader for Windows — Grabbit How Talkie OCR Helps Vision-Impaired & Dyslexic Users Read the World Around Them VRCFaceTracking安装和iPhone面捕配置教程,有bug Even CrowdStrike Can't See Your Agents The Automation Gold Rush: What n8n Workflows and Claude Are Opening Up for Developers Right Now
I built an MCP server that charges AI agents per call using x402 micropayments
Toolstem Team · 2026-06-16 · via DEV Community

I built an MCP server that charges AI agents per call using x402 micropayments

By Toolstem


You built an MCP server. Agents call it. You pay the API costs. They pay nothing. Every tools/call invocation burns a few cents of your budget while the agent that triggered it has no idea — and no incentive to care. The standard answer is "wrap it in a SaaS subscription," but that requires a human in the loop to sign up, enter a card, and manage a billing relationship. Most agents don't have humans watching every call. What if the agent just paid for its own data, automatically, without a human anywhere in the loop?

That's what I built. Here's what I learned.


What x402 actually is

HTTP 402 — "Payment Required" — has existed since 1996. It was reserved for future use and then essentially forgotten for 28 years. The x402 protocol revives it as a machine-to-machine payment standard built on EIP-3009.

The flow looks like this:

  1. Agent hits your endpoint with a normal request
  2. Server returns HTTP 402 with a JSON body: payment amount, recipient address, network, token, nonce
  3. Agent signs an EIP-3009 transferWithAuthorization — a pre-authorized USDC transfer that the server can submit on behalf of the agent
  4. Agent retries the request with an X-Payment header containing the signed authorization
  5. Server verifies the signature on-chain, submits the transfer, and responds with data
  6. Settlement confirmed in under 2 seconds on Base mainnet; gas is roughly $0.001

No Stripe account. No OAuth flow. No human sign-up. No waiting for a monthly invoice. The agent pays exactly what it used, settled atomically, without ever needing ETH for gas (the server submits the transfer, so only the server needs a small ETH float for gas — the agent only needs USDC).

This is the "HTTP + payment" primitive that should have existed all along. It's pre-product-market-fit, but the underlying mechanic is sound.


The architecture

The core infrastructure is a Cloudflare Worker acting as a paywall proxy. It sits between the public internet and your upstream MCP server, intercepting every request before forwarding it.

Agent → Cloudflare Worker (x402 paywall) → Upstream MCP Server → Data sources

The Worker handles the payment lifecycle:

  • Receives request
  • Checks for a valid X-Payment header
  • If absent: returns 402 with payment details
  • If present: verifies the EIP-3009 signature, checks nonce freshness and amount, submits the on-chain transfer
  • Forwards to upstream MCP only on successful verification

One non-obvious constraint: initialize and tools/list must be free. Agents need to discover your tool surface before they can decide whether to pay. If you charge for discovery, you never get called at all. Any directory health-check probe (Glama, mcp.so, PulseMCP) also expects a free tools/list response — paywall those and you fall off the listings. The Cloudflare Worker routes those two MCP message types to an unmetered path.

Per-tool pricing is where the model gets interesting. Our SEC EDGAR server runs three tiers:

Tool Tier Per call
getCompanyFilingsSummary Cheap $0.005
getInsiderSignal Standard $0.05
getInstitutionalSignal Standard $0.05
getMaterialEventsDigest Premium $0.50
compareDisclosureSignals Premium $0.50

A filings summary is one EDGAR lookup. compareDisclosureSignals cross-references insider trades, 13F institutional moves, and 8-K clusters across multiple companies. The compute difference is roughly 100x — so the price difference is 100x. Flat pricing treats them identically, which means you're subsidizing every expensive call with revenue from cheap ones.

The Finance MCP server (our first server) launched at a flat $0.005 across all tools. Six weeks in, we had $0 revenue. That table above is the lesson.


Code: agent side

The LangChain integration wraps the payment layer into a custom fetch function. Everything else is standard agent code.

TypeScript / LangChain.js:

import { createReactAgent } from "@langchain/langgraph/prebuilt";
import { createFinanceTools } from "langchain-toolstem/finance";
import { createX402Fetch } from "langchain-toolstem/x402";

// fetch that auto-signs USDC payments on HTTP 402
const fetchPay = await createX402Fetch({
  privateKey: process.env.X402_PRIVATE_KEY!,
  maxPaymentUsd: 0.05, // per-call safety cap
});

// discovers tools live via MCP tools/list (free)
const tools = await createFinanceTools({ fetch: fetchPay });

const agent = createReactAgent({ llm, tools });
// each tools/call costs $0.01 USDC, settled on Base mainnet

The createX402Fetch wrapper intercepts any 402 response, signs the EIP-3009 authorization using the agent's private key, and retries transparently. The maxPaymentUsd cap is a safety rail — the agent refuses to sign any authorization above that threshold, protecting against a misconfigured or malicious server quoting an unexpected price.

Tools are discovered live via tools/list on startup (free, not metered), so the agent always sees the current tool surface without anything hardcoded into the package.

Python / LangChain:

from langchain_toolstem import create_finance_tools, create_x402_httpx_client

client = create_x402_httpx_client("0xYOUR_PRIVATE_KEY")
tools = await create_finance_tools(client=client)
# same pattern — agent signs payments, settles on Base

For environments where you need a plain http:// URL rather than a custom fetch (e.g., MultiServerMCPClient), createX402Proxy spawns a local reverse proxy at localhost:4021 that handles payment signing, so you can point any standard MCP client at it directly.

The package is langchain-toolstem on npm (~1,400 installs/month) and langchain-toolstem on PyPI.


What we learned shipping this

Flat pricing was wrong. The Finance server ran flat $0.005 per call for six weeks. Zero meaningful revenue. The math is brutal: even if you hit the plateau of ~1,400 monthly active users typical for popular Apify actors, $0.005 flat yields ~$7/month. The ceiling is too low for a server with real API costs underneath it. Tiered pricing maps price to actual compute — it was the right call, and we should have done it from day one.

The free discovery layer is non-negotiable. Agents must be able to call tools/list without triggering a payment. This isn't just good UX — it's a technical requirement for every MCP directory that health-checks your listing. We had one bug early on where an unbilled discovery path was missing, and the server briefly vanished from Glama's listings because the probe got a 402 instead of a tools manifest.

The heartbeat problem is real. Coinbase's x402 Bazaar delists your endpoint if it goes 30 days without a confirmed paid call. For a server that's still building an audience, that's a serious risk — you could drop off the directory just as someone is about to discover you. We solved this with a GitHub Actions cron that runs every 6 hours, makes a real paid call from a funded heartbeat wallet, and posts the result to a log. Cost: roughly $0.06/day in USDC. Insurance against delisting: worth it.

The heartbeat revealed another bug: our initial implementation was calling tools/call without first completing the MCP initialize handshake. The server was rejecting the call silently. Fixing it required tracing the raw MCP message sequence. A reminder that the protocol has state — it's not a stateless REST API.

One external paying customer. $0.01. I want to be direct about this. The payment rail works end-to-end — 49 self-test transactions confirmed on Base mainnet, and one external payment from wallet 0x9CC4 on 2026-06-03 for $0.01. That's the full external revenue picture. The mechanism is proven; adoption is the open question.

We're self-audited on the security side (no third-party audit yet), and we've run independent AI reviewer passes on the codebase. The x402 verification logic, nonce handling, and private key isolation are the parts that matter most and got the most scrutiny.


Where this is going

x402 is a protocol looking for a distribution moment. The primitive is correct: HTTP-native, cryptographic, machine-to-machine, no humans required. What it lacks is the agent wallet layer becoming standard.

That moment is approaching. Coinbase AgentKit and CDP wallets are building the infrastructure for agents to hold and spend funds natively. When agent wallets are as common as agent LLM clients, x402 becomes the obvious way to monetize any agent-facing API — a single fetch wrapper is all the integration a developer needs.

Until then, the pattern is in early-adopter territory. The agents that can use it today are ones whose operators have explicitly provisioned a funded wallet and a private key in their environment. That's not zero — it's the population of developers actively building agentic systems who care about cost attribution.

The server is live at mcp.toolstem.com. Try it with a funded Base wallet. initialize and tools/list are always free — you can inspect the full tool surface before committing a cent.


The Finance and SEC EDGAR MCP servers are listed in Coinbase's x402 Bazaar. Source: github.com/toolstem/toolstem-mcp-server and github.com/toolstem/toolstem-sec-mcp-server. Walletless demo at toolstem.com/playground.