LlamaIndex TS

Install

npm install llamaindex @llamaindex/workflow @llamaindex/openai agent402-llamaindex @x402/fetch @x402/core @x402/evm viem

Example

import { agent } from "@llamaindex/workflow";
import { openai } from "@llamaindex/openai";
import { agent402Tools } from "agent402-llamaindex";
import { wrapFetchWithPayment } from "@x402/fetch";
import { x402Client } from "@x402/core/client";
import { registerExactEvmScheme } from "@x402/evm/exact/client";
import { privateKeyToAccount } from "viem/accounts";

const payClient = new x402Client();
registerExactEvmScheme(payClient, { signer: privateKeyToAccount(process.env.AGENT_KEY) });
const payFetch = wrapFetchWithPayment(fetch, payClient);

// web search, then a cited answer: both paid per call from the wallet
const { tools } = await agent402Tools({ slugs: ["search", "answer"], freeOnly: false, fetch: payFetch });

const myAgent = agent({ tools, llm: openai({ model: "gpt-4o-mini" }) });
const res = await myAgent.run("Search the web for x402 payment protocol adoption, then answer with citations: what is the x402 payment protocol?");
console.log(res.data.result);

No wallet yet? The pure-CPU tools (hash, uuid, base64, markdown, JSON and more) run free with proof-of-work and no wallet:

// free tier: every tool listed here is compute-payable (proof-of-work, no wallet)
const { tools } = await agent402Tools({ slugs: ["hash", "markdown-to-html", "text-stats", "uuid"] });

What it exposes

• One native tool per catalog slug you pass in slugs (a short list gives the model better tool selection).
• With no slugs, the free-tier catalog: every compute-payable tool (freeOnly: true is the default).
• An execute(name, args) function that runs a tool call and pays for it underneath.
• baseUrl points the adapter at a self-hosted Agent402 instance.

How payment works

• Free tier: compute-payable tools settle with a sha256 proof-of-work solved in-process. No wallet, no API key.
• Wallet-only tools (browser, network, memory, live data): set freeOnly: false and pass a payment-wrapped fetch. An @x402/fetch fetch pays in USDC over x402; a stock mppx fetch pays over MPP. The same 402 carries both offers.
• Prices come from the live catalog at /api/pricing and are quoted in every 402 before anything is signed.

A call that fails is not charged: settlement runs after the tool and only on a successful response. Background: What x402 is · MPP (Machine Payments Protocol).

Tools to try first

Every tool, price and input schema: the catalog.

Links

agent402-llamaindex on npm · source on GitHub · configuration reference
Guides: Make your AI agent pay for what it needs: x402 in 5 minutes · Durable memory for AI agents - no accounts, the wallet is the identity
All integrations