Agent402 MCP Server

Agent402: 500+ tools, metered models and finished reports for AI agents

Live at agent402.tools.

The paid door for AI agents: 500+ tools, metered models and finished reports, paid per call in USDC over x402 and MPP, or by card. The applied layer of Agentic Finance. Open source, self-hostable, MCP-native.

Operated by Havok Holdings LLC · Live · Why pay here · Receipts · Status · Security · Company

Try it in 30 seconds: claude mcp add agent402 -- npx -y agent402-mcp, or paste https://agent402.tools/mcp into any MCP client. The free tier needs no wallet. Verified setup blocks for Cursor, VS Code, Windsurf, Cline, Roo Code, Codex CLI, Gemini CLI, Continue, ElizaOS and AgentCore: agent402.tools/guides/agent-hosts (shortlinks agent402.sh/<host>, e.g. agent402.sh/cursor).

Who pays: agents pay per call (from $0.001; models metered under a quoted ceiling and settled at actual usage); people buy finished reports ($2 to $5) and monitors ($5 a month) by card; sites charge crawlers with the tollbooth.

What Agent402 is

Agentic Finance is software agents paying and getting paid on their own: discovering a service, paying per request from a wallet over open protocols (x402, MPP), receiving a verifiable receipt, and earning per request in return. Agent402 is its applied layer - explainer: agent402.tools/agentic-finance · glossary.

Selling an x402 API? List it on the index for free at /sell (POST /api/index/register: self-serve, no signup, 0% take). The router sends buyers to a seller once its routes show settled payments from several independent buyers; calls priced at $0.01 or less can be tried before that, marked unproven on the buyer's receipt. Charging crawlers on your own site instead? agent402-tollbooth puts a 402 gate in front of it.

What makes it different: Agent402 is open-source and self-hostable - and a single integration gives a buyer three free primitives over the whole x402 ecosystem:

• Find - /api/find?q={task} resolves a task description to the best-matching tools (route, price, schema, ready example).
• Route - POST /api/route is the neutral Smart Order Router: rank tools across every x402 seller crawled (auto-discovered from the Coinbase CDP Bazaar), health-aware, with include=external to exclude us. The response returns the top N and carries matched for how many scored.
• Leaderboard - GET /api/leaderboard is the public on-chain ranking of x402 sellers by Base USDC settled volume - calls served, totalUsd, unique buyers per seller. It returns the top N, never the whole board (25 by default, 50 the ceiling; totalSellers carries the full count). Pipeline: Bazaar → eth_getLogs → per-call ceiling → aggregate by payTo. Hourly snapshot.

Plus the catalog - 500+ strong: search/answer as the MCP front door, then 500+ tools and curated skill packs (multi-tool workflows callable as MCP prompts) - all runnable yourself, plus agent402-tollbooth - an open pay-per-crawl gate for the other side of x402.

Two doors, two price lists. Agents pay per call in USDC (x402 / MPP) or free via proof-of-work, and a finished report costs $0.60 to $2.00 that way. People pay by card: the same reports at agent402.tools/reports (company dossier, 13F fund report, insider flow, market brief, deep research, FDA recall, domain audit) are $2 to $5, monitors that re-run on change at /monitors are $5 a month, and a prepaid credits key already issued keeps paying per successful call (Authorization: Bearer a402_…; new credits are not on sale). The card price includes payment processing. An agent paying per call pays the lower tool price for the same report.

Two payment wires, one URL. Every paid endpoint accepts x402 (PAYMENT-SIGNATURE, USDC on 12 chains or USDG on Robinhood Chain) and MPP (Machine Payments Protocol, the IETF-track Payment HTTP auth scheme co-authored by Tempo and Stripe): the same 402 carries both, an mppx client pays out of the box, and MPP settles on Base and Celo (USDC), natively on Tempo (USDC.e or PathUSD via Tempo's relay), or by card over MPP (Stripe stripe/charge, offered on routes priced $0.50 and up when the operator configures it). The MPP marketplace lists every MPP seller we can verify live. Details: What is MPP · live MPP settlements.

Framework adapters (drop-in tools for the major agent stacks - auto-payment underneath):

500+ strong - live web search and cited answers as the MCP front door, then ready-to-use web tools and multi-tool skill packs for your AI agent, from one server. Every one priced per call and settled on-chain. Browser rendering, web search, PDFs, images, OCR, live financial/crypto/macro data, SEC EDGAR, deterministic stats, forecasting, and options/bond pricing (Black-Scholes, YTM), compression, and 150+ pure-CPU utilities. Run it yourself for free in 30 seconds (MCP or plain HTTP, no API keys, no signup - the free tier and x402/MPP payments never need a key; only a prepaid credits key already issued is a bearer key), connect it to Claude/ChatGPT/any MCP client, and add your own tools in a few lines. No language model writes a utility tool's answer, and every tool CI can run without a third-party key is re-tested against its own example before every release. The model-backed surfaces are explicit and priced as such: the /v1 gateway (metered under a quoted ceiling, or flat tiers) and the finished report products (/v1/research, /v1/dossier, ... below).

Optionally, the same server can charge per call over the x402 protocol (USDC on Base, Solana, Polygon, Arbitrum, Monad, Celo, Avalanche, Sei, Optimism, Arc, Stellar & Algorand, plus USDG on Robinhood Chain - 13 chains) - so the instance you self-host for free can also be a hosted, monetized one. Free mode is one setting (FREE_MODE=true); a paid instance needs a payout wallet.

🟢 Hosted demo: agent402.tools · 📖 Wiki · 📦 npm · 🔌 MCP Registry · 🧩 Smithery

Run it yourself in 30 seconds

Pick whichever fits. They are all free and need no wallet:

1. Zero install - add the hosted connector to Claude (claude.ai → Settings → Connectors → Add custom connector):

https://agent402.tools/mcp

2. One command - run the MCP server locally (the pure-CPU tools work with no key; it pays the tiny proof-of-work for you):

npx -y agent402-mcp
# in Claude Code:  claude mcp add agent402 -- npx -y agent402-mcp

3. Clone and host the whole thing (all 500+ tools as an HTTP API + MCP, free mode, no payments):

git clone https://github.com/MikeyPetrillo/Agent402 && cd Agent402
npm install
FREE_MODE=true npm start          # → http://localhost:3000  (HTTP API + /mcp)

# try a tool over HTTP - no auth in free mode
curl -s -X POST localhost:3000/api/hash -H 'content-type: application/json' \
  -d '{"text":"hello world","algo":"sha256"}'

4. One-click deploy to Railway (full self-hosted instance - adds optional Postgres + Redis plugins for analytics + response caching):

Boots straight from the repo's Dockerfile. Optional plugins are auto-detected via env: add Redis → REDIS_URL enables the upstream response cache (X-Cache: hit|miss), add Postgres → DATABASE_URL enables the public /api/analytics dashboard. No env vars required to boot in free mode.

Pay per call from a wallet

Every paid route answers 402 with both an x402 offer and MPP challenges, so either client pays it. No API key, no signup: the wallet is the account.

MPP (USDC.e or PathUSD on Tempo, listed first in the 402, or USDC on Base or Celo) with mppx:

import { Mppx, tempo, evm } from "mppx/client";
const account = (await import("viem/accounts")).privateKeyToAccount(process.env.AGENT_KEY);
const res = await Mppx.create({ methods: [tempo.charge({ account }), evm.charge({ account })] }).fetch("https://agent402.tools/api/uuid");

x402 (USDC on Base and other chains) with @x402/fetch:

import { wrapFetchWithPayment } from "@x402/fetch"; import { x402Client } from "@x402/core/client"; import { registerExactEvmScheme } from "@x402/evm/exact/client";
const client = new x402Client(); registerExactEvmScheme(client, { signer: (await import("viem/accounts")).privateKeyToAccount(process.env.AGENT_KEY) });
const res = await wrapFetchWithPayment(fetch, client)("https://agent402.tools/api/uuid");

More: Paying with MPP · Paying with x402.

What's in the catalog (500+ tools)

Every tool earns its place: priced to market and settled on-chain, and every tool CI can run without a third-party key is tested against its own example on every CI run. CI holds a 400-entry catalog floor and verifies the “500+” claim against the running catalog (scripts/sync-count.js --check). The catalog grows only when a tool is worth calling.

Full schemas live in /openapi.json; a machine-readable catalog is at /api/pricing and /llms.txt. Don't know which tool you need? /api/find?q=<task> resolves a task description to the right tool - route, price, schema, and a ready example - so an agent skips the token-heavy "search around to find a tool" step.

LLM gateway (/v1) - chat, embeddings, rerank, images, speech & finished reports, pay per call

Point any OpenAI SDK at base_url = https://agent402.tools/v1 and pay per call in USDC over x402 or MPP - no API key, no signup, no account (a prepaid credits key already issued also works, see For humans):

Streaming (stream: true), full tools/function-calling passthrough, an opt-in prompt cache on the chat tiers (cache: true → byte-identical repeats free for 10 minutes; the repeat's payment is not charged, except a Tempo transfer sent before the call, which is final when sent), upstream failover chains that end in a canary-proven model, and a free GET /v1/models listing every model with its tier and caps. A real-money canary buys the core surfaces daily (the chat tiers, the messages and responses wires, embeddings, rerank, image generation and speech) - streaming, routing disclosure, and both cache behaviors included.

Finished report products live on the same /v1 prefix: one paid call returns a complete, cited report (JSON, with a sources appendix) rather than a raw model turn. Every report is grounded in primary data the server fetches itself (EDGAR filings, openFDA, DNS/TLS probes, on-chain reads, a live web search) before synthesis; ipo-report is fully deterministic (no model at all):

Reports are wallet-only (x402 / MPP / prepaid credits, never proof-of-work) and take a few minutes to generate. The same reports are sold to people by card at agent402.tools/reports for $2 to $5 depending on depth. The card price includes payment processing; an agent paying per call over x402 or MPP pays the lower tool price for the same report.

Two companion tools close the loop: POST /api/route/execute ($0.01, with execute-plus $0.05, execute-max $0.55 and execute-pro $3.30 tiers for pricier tools) resolves a task description to the best tool and runs it in one paid call, including, with include:"external", tools sold by other x402 sellers (it routes on proven on-chain settled volume, with one exception - a Base or Solana seller with no history yet is tried only after every proven candidate, under a small per-call ceiling, and flagged unproven on the receipt - buys the result from them on the chain you paid on - Base, Solana, Algorand or Tempo - and sells it to you; see the Smart Order Router guide), and POST /api/my-usage ($0.005) returns the paying wallet's own purchase history - no wallet parameter; the x402 payment is the identity, so nobody can read another wallet's profile.

For humans: reports and monitors

The same catalog has a card-paying front door (Stripe Checkout; the operator enables it with STRIPE_SECRET_KEY, otherwise these pages simply do not mount):

An issued credits key is understood by the SDKs: agent402-mcp reads AGENT402_CREDITS_KEY and agent402-client takes { creditsKey }.

Decide - describe the job, get the plan

POST /api/decide turns a plain-language task into a call-ready plan: which tools, from this catalog and from outside x402 sellers whose 402 was verified in the last week, in what order, with fallbacks, inputs that validate against each tool's schema, and cost and latency estimates. Priced by depth: quick (one best tool), plan (steps and fallbacks) and full (plus a compiled prompt), from $0.005 per decision over x402 or MPP.

curl -X POST https://agent402.tools/api/decide \
  -H "Content-Type: application/json" \
  -d '{"task":"Check whether wallet 0xd8dA...6045 is on a sanctions list, then list its recent token transfers on Base","depth":"plan"}'

• Neutral ranking. One formula scores every candidate (fit, observed reliability, price, schema quality, freshness) with no term for who sells the tool, and every step says whether the tool is ours (firstParty).
• The fee comes back. The decision fee is returned as a 24-hour credit toward POST /api/decide/execute, which runs the plan: our tools directly, outside steps bought from the seller and resold at the seller's price plus a disclosed 5% markup. Spend stops at the budget, fallbacks are tried in order, and a run where no step succeeds is not charged.
• On MCP the same pair is decide.plan and decide.execute; decide.feedback (free) records whether a step worked. Page: agent402.tools/decide.

Skill packs - 70+ multi-tool workflows

For jobs that span several tools - "audit a domain", "diagnose deliverability", "work up a time-series", "peel an opaque blob" - Agent402 ships curated skill packs: ordered, typed sequences of tool calls with a Claude-ready prompt template. Callable as MCP prompts (prompts/list → prompts/get { name, arguments }) or plain HTTP at GET /api/skill-packs/{slug}/prompt (every slug is listed in the free JSON index at /api/skill-packs.json). A task-shaped query to catalog.search (the hosted connector's search tool; the older search_tools name still works as an alias) returns the matching pack alongside individual tools.

All 80+ packs at /skills · JSON index at /api/skill-packs.json · on MCP the packs appear under prompts/list so any MCP-aware client picks them up automatically.

x402 Index - Find · Route · Leaderboard

Agent402 is also the open routing + ranking layer for the whole x402 ecosystem: it crawls public x402 sellers (the local catalog + an auto-discovered set from the Coinbase CDP Bazaar, refreshed hourly) and exposes them through three free surfaces - same logic as /api/find: discovery primitives shouldn't cost money.

# "I need an OCR tool - find me the cheapest healthy one anywhere on x402"
curl -X POST https://agent402.tools/api/route \
  -H 'content-type: application/json' \
  -d '{"query":"ocr image to text","top":5}'

# "Who are the most-used x402 sellers right now? (on-chain proof, not self-reports)"
curl 'https://agent402.tools/api/leaderboard?top=25&include=external'

Dispatch is labelled, never implied: every /api/route row and /api/index seller carries routerDispatchEligible and routerDispatchReason (crawl_failed, network_unknown, settlement_required, settlement_checked_at_pay_time, eligible, ...); executeVia appears only on a row the router will pay right now (executeViaCallableNow: true), otherwise the tier moves to executeViaWhenEligible. routable is crawl readiness, never a promise to pay. A manifest-priced route is also read live once and then weekly, so the chains its 402 actually offers reach the row even when the seller's manifest lags.

Health-aware: sellers whose last few crawls errored are excluded from the router (a buyer routed to a dead seller wastes money). Healthier sellers also break ties at equal match score and price, so flaky-but-cheap sellers lose to reliable ones. Brand-new sellers (no history yet) get the benefit of the doubt.

Judged fit: the lexical score rewards a listing whose name repeats the task's words, so one seller's twenty narrow routes could out-rank a general tool that does the job. A judgment model now reads a shortlist - at most two rows per seller, plus the best matches from the local catalog - and moves the listing that actually does the task to the top; when several do it equally well, the cheapest wins. It never adds a seller the settlement gate did not admit, it can say "none of these" (route-execute then refuses before anything is paid), and it fails open to the lexical order. The tools themselves still run no model; this is only how the router chooses.

Operators get 3-rail attribution on the dashboard (/api/stats, /__operator, token-gated - it answers 404 without the operator token): USDC vs. proof-of-work vs. heartbeat-probe traffic are counted separately - and the heartbeat rail is gated on a POW_SECRET-signed token (not a spoofable User-Agent), so the operator view reflects real external demand.

For API sellers: /sell is the front door - list an existing x402-speaking API on the index for free with POST /api/index/register (health-routed, 0% take, self-serve, no signup), or install agent402-tollbooth to put a pay-per-crawl gate in front of a site that isn't x402-native yet.

From code, the agent402-client npm package wraps all of this - find() a tool, then call() it, paying automatically (a built-in proof-of-work for free tools, your x402 wallet for paid ones), with caching and idempotent retries:

npm install agent402-client

import { Agent402 } from "agent402-client";
const a = new Agent402();                       // free tier (proof-of-work)
const out = await a.call("hash", { text: "hello world", algo: "sha256" });

// a prepaid credits key already issued pays wallet-only tools (new credits are not on sale)
const b = new Agent402({ creditsKey: "a402_..." });

Documentation truncated — see the full README on GitHub.