AgentScout MCP Server
OfficialClick on "Install Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@AgentScout MCP Serverread https://example.com to get the article as markdown"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
AgentScout
Open-source clients for AgentScout — an agent-native web read/extract/crawl service paid
per fetch over x402. Give it a URL and get back clean markdown (read),
structured JSON validated against your schema (extract), a free price probe (quote),
or a whole-site crawl — every paid fetch settling in USDC on Base, with no signup and no
API keys.
There are two ways to pay, auto-detected by the client:
Wallet-as-payer (the default): a signable EVM wallet pays each fetch inline via x402. AgentScout mints and manages a local wallet on first use, so an agent "just works" once that wallet is funded.
Account-key (for managed wallets that can't sign — e.g. awal): an opaque
ak_…bearer token identifies the account and debits prepaid credits. Credits are funded out-of-band via AgentKV, so any signing wallet can fund the account and fetches carry only the bearer.
Scout content is not encrypted — it is public web data by definition (unlike AgentKV, which encrypts values client-side). A scout response is plaintext; do not treat it as private.
This repository holds the client surface — the SDK, CLI, MCP server, and Claude plugin. The AgentScout service (the backend) is operated separately; these clients talk to it over the public x402 + EIP-712 protocol.
Packages
Path | Package | What |
| TypeScript SDK — sign + pay + fetch | |
| the | |
— | Claude Code plugin (wraps the MCP server) |
Related MCP server: JMT x402 MCP Server
npm scopes
Two npm scopes separate the platform from the service:
@agentx402-ai/*— the platform scope:@agentx402-ai/core, a shared SDK for auth, payment, usage tracking, error handling, and retry logic, consumed by every agentx402 service. It lives in its own repo (agentx402-ai/core) and is a published dependency of the packages here.@agentscout/*— the scout service scope:@agentscout/clientand@agentscout/cli(this repo), which depend on@agentx402-ai/corefor shared plumbing.
Keeping @agentx402-ai/core in its own repo lets sibling services (e.g. @agentkv/client)
share it without depending on the AgentScout repo.
Quick start (SDK)
npm install @agentscout/clientimport { AgentScout } from "@agentscout/client";
import { privateKeyToAccount } from "viem/accounts";
const scout = new AgentScout({
signer: privateKeyToAccount(process.env.PRIVATE_KEY as `0x${string}`),
endpoint: "https://api.agentx402.ai",
maxSpendUsd: 0.05, // optional: refuse any single fetch over $0.05
});
// quote → FREE price probe (never signs, never spends)
const price = await scout.quote("https://example.com");
// read → clean markdown (paid per fetch in USDC via x402)
const { markdown, title } = await scout.read("https://example.com");
// extract → JSON validated against a JSON Schema you supply
const { data } = await scout.extract("https://example.com", {
type: "object",
properties: { headline: { type: "string" } },
});
// crawl → walk a site up to max_pages, get a results manifest
const result = await scout.crawl("https://example.com", { maxPages: 10 });CLI
npm install -g @agentscout/cli
export AGENTSCOUT_PRIVATE_KEY=0x... # endpoint defaults to https://api.agentx402.ai
agentscout quote https://example.com # free price probe
agentscout read https://example.com # → clean markdown (paid per fetch)
agentscout extract https://example.com --schema ./schema.json
agentscout crawl https://example.com --max-pages 10No wallet? Leave AGENTSCOUT_PRIVATE_KEY unset and AgentScout mints and manages a local wallet
on first use (a 0600 keystore under ~/.agentscout), printing its address — fund that address
with USDC on Base, then retry. Cap spend any time with AGENTSCOUT_MAX_SPEND_USD (per fetch) and
AGENTSCOUT_MAX_SESSION_SPEND_USD (cumulative); a malformed value fails closed.
Account-key mode (works with awal / any managed wallet)
For a managed wallet that can't sign (e.g. awal), use an account key funded out-of-band via AgentKV; every fetch then carries only the bearer and debits the account's prepaid credits:
export AGENTSCOUT_ACCOUNT_KEY=ak_... # a bearer minted + funded via AgentKV
agentscout read https://example.com # debits the account's prepaid creditsThe client auto-selects account-key mode when AGENTSCOUT_ACCOUNT_KEY is set (or a stored
account key exists and no AGENTSCOUT_PRIVATE_KEY is set); otherwise it uses the wallet.
Publisher tolls (--max-toll-usd) are wallet-mode only — an ak_ caller cannot front a
real-USDC toll, so setting one in account-key mode fails fast (tolls_require_x402) before any
request is issued.
MCP server / Claude plugin
agentscout mcp exposes the service as MCP tools — scout_read, scout_extract, and
scout_crawl (the three paid fetch verbs), plus the free scout_quote (price a fetch before
spending) and scout_crawl_status (resume a long crawl by jobId) — for Claude Desktop / Code /
Cursor. The paid verbs are annotated as state-changing (never readOnlyHint) so a client knows
to prompt a human before spending.
The plugin/ directory packages this as an installable Claude Code plugin. In
Claude Code:
/plugin marketplace add agentx402-ai/claude-plugins
/plugin install agentscout@agentx402Claude Code then prompts for your wallet private key (stored in your OS keychain) and the
optional AgentScout endpoint (defaults to the hosted service), and auto-starts the MCP server —
verify with /mcp. Full steps: plugin/README.md.
How it works
Pay per fetch over x402. Each paid verb (
read,extract,crawl) is priced by the server's402challenge and settled in USDC on Base via x402 (EIP-3009transferWithAuthorization). The SDK signs the challenge's exact quoted amount — never a self-computed sum — pinning the network, the canonical USDC token, and (when you setexpectedPayTo) the recipient before signing.quoteis free: it prices a URL without ever signing or spending.Wallet-as-payer, or account-key credits. In wallet mode a signable wallet pays each fetch inline and is itself the identity. In account-key mode an opaque
ak_…bearer is the identity and fetches debit prepaid credits funded out-of-band via AgentKV — so any signing wallet can fund the account, decoupled from the fetches.Publisher tolls. Some publishers charge a per-fetch toll on top of the base price; the SDK pays it only up to your
maxTollUsdceiling, and only in wallet mode (anak_caller cannot front a real-USDC toll, so tolls are refused client-side in account-key mode).Client-side spend caps.
maxSpendUsd(per call) andmaxSessionSpendUsd(cumulative) refuse — never silently cap — any op that would exceed them, checked before the challenge is signed.No encryption. Scout returns public web content in the clear; there is no encryption key and nothing zero-knowledge about a response (unlike AgentKV).
License
This server cannot be installed
Maintenance
Resources
Unclaimed servers have limited discoverability.
Looking for Admin?
If you are the server author, to access and configure the admin panel.
Related MCP Servers
- AlicenseAqualityAmaintenanceContext-efficient MCP server for AI agents. fetch_extract reduces token usage by a median 98.1% (53,820 → 2,001 tokens) vs raw HTML — saves ~$0.156/call at Sonnet pricing. Pay $0.02 USDC on Base via x402. First call free per wallet. 10 always-free utility tools included.Last updated331MIT
- AlicenseBqualityCmaintenanceExposes 25 paid API endpoints as MCP tools for AI agents, with payments in USDC on Base mainnet via the x402 protocol, enabling tasks like web search, company intelligence, and crypto research.Last updated2573MIT
- Alicense-qualityCmaintenancePay-per-call MCP server for WebberSites x402 Data API, offering 45 tools for AI agents: web scraping, document extraction, SEO audits, linting, crypto data, and more, with payment via USDC on Base.Last updated26MIT

fetcher.cloudofficial
Alicense-qualityBmaintenanceMCP server that gives AI agents web capabilities such as rendering pages, extracting content, resolving links, and inspecting domains, with per-call USDC payments via x402.Last updatedMIT
Related MCP Connectors
Pay-per-call web scraping for AI agents via x402 on Base USDC. Six tools, no signup.
32 paid x402 endpoints for crypto, Zora & on-chain analysis. 10 MCP tools. USDC on Base.
MCP marketplace: agents pay per call in USDC via x402. Plus Base chain data and a USDC<->bank ramp.
Latest Blog Posts
- Who's Calling? MCP Hosts Are an Identity Blind Spot (And the Spec Knows It)By Om-Shree-0709 on .mcpAgent IdentityOAuth 2.1
- Your AI Chatbot Just Exposed Your CEO's Salary to an InternBy Om-Shree-0709 on .Agent IdentityMCP SecurityOAuth Delegation
- Why MCP Servers Need Execution Sandboxing (And Why Your Current Stack Isn't Enough)By Om-Shree-0709 on .Agentic AiPrompt InjectionWebAssembly
MCP directory API
We provide all the information about MCP servers via our MCP API.
curl -X GET 'https://glama.ai/api/mcp/v1/servers/agentx402-ai/agentscout'
If you have feedback or need assistance with the MCP directory API, please join our Discord server