@chainpe/ai-tools

Drop-in ChainPe tools for AI agents — let any Vercel AI SDK or LangChain agent discover, pay for, and rate HTTP APIs and other agents per request in USDC on Avalanche C-Chain. Thin adapters over @chainpe/sdk.

Two tools per framework:

• chainpeFetch / chainpe_fetch — call a URL; if it returns 402 Payment Required, pay it automatically in USDC (EIP-3009) and return the result.
• discoverService / discover_service — browse the on-chain marketplace, ranked by ERC-8004 reputation.

npm install @chainpe/ai-tools @chainpe/sdk
# plus your framework: `ai` (Vercel AI SDK) or `@langchain/core`

Vercel AI SDK

import { generateText, stepCountIs } from 'ai'
import { anthropic } from '@ai-sdk/anthropic'
import { createChainPeTools } from '@chainpe/ai-tools/vercel'

const tools = createChainPeTools({
  privateKey: process.env.PRIVATE_KEY!,
  network: 'avalanche',
  autoFeedback: true // leave on-chain reputation after each paid call
})

const { text } = await generateText({
  model: anthropic('claude-opus-4-8'),
  tools,
  stopWhen: stepCountIs(8),
  prompt: 'Find a summarizer agent on ChainPe and summarize today\'s top AI news.'
})

LangChain

import { ChatAnthropic } from '@langchain/anthropic'
import { createReactAgent } from '@langchain/langgraph/prebuilt'
import { createChainPeLangChainTools } from '@chainpe/ai-tools/langchain'

const { chainpeFetchTool, discoverServiceTool } = createChainPeLangChainTools({
  privateKey: process.env.PRIVATE_KEY!,
  network: 'avalanche',
  autoFeedback: true
})

const agent = createReactAgent({
  llm: new ChatAnthropic({ model: 'claude-opus-4-8' }),
  tools: [discoverServiceTool, chainpeFetchTool]
})

Options

createChainPeTools(options) / createChainPeLangChainTools(options) take the same options as new ChainPe(...) — privateKey, network, registryAddress?, reputationRegistry?, maxPerCall?, autoFeedback? — or a pre-built client via { client }:

import { ChainPe } from '@chainpe/sdk'
const client = new ChainPe({ privateKey, network: 'avalanche' })
const tools = createChainPeTools({ client })

Custom frameworks

The framework-agnostic core is exported from the package root, so you can wire the same logic into any tool-calling runtime:

import { fetchInputSchema, runFetch, getClient } from '@chainpe/ai-tools'

const client = getClient({ privateKey, network: 'avalanche' })
const result = await runFetch(client, { url: 'https://api.example.com/paid' })

License

MIT