dyoe-agent-tools-mcp
# DYOE Agent Tools โ MCP Server
Pay-per-call tools for AI agents, settled in **USDC on Base** via the [x402](https://www.x402.org) protocol. Trust checks, due-diligence, commerce signals, market data โ plus the one thing nobody else sells: a **human-verified approval layer**.
> **The trust layer of the agent economy.** In a world of anonymous autonomous agents, intelligence is commodity and *trust* is the scarcity. These tools let an agent verify who it's about to trust or pay โ and, when it matters, get a **real human** to sign off.
Live API: **https://agents.dyoeway.org** ยท Discovery manifest: [`/.well-known/x402.json`](https://agents.dyoeway.org/.well-known/x402.json)
๐ **Open protocol:** [Credential Spec](./SPEC.md) โ verify any DYOE credential yourself, no permission needed.
๐ **How we decide trust:** [Trust Methodology](./TRUST-METHODOLOGY.md) โ our scoring, approvals, and how we handle mistakes.
## What's inside (16 tools)
| Tool | Price | What it does |
|---|---|---|
| `trust_check` | $0.01 | Scam/fraud/phishing/rug-risk verdict on any site, business, wallet or endpoint |
| `guardian_approve` | $3.00 | **Human-verified** approval of an agent action (a real person signs off) |
| `intel_dossier` | $0.10 | Full due-diligence dossier on a company/site in one call |
| `site_audit` | $0.05 | Trust + SEO + contactability audit, scored 0โ100 |
| `merchant_trust` | $0.05 | Is this online store safe to buy from? |
| `token_intel` | $0.05 | Trust-vetted token data โ price **+** scam/rug read on the project |
| `extract_text` ยท `extract_contacts` ยท `page_metadata` | $0.01 | Web extraction for research & enrichment agents |
| `crypto_price` ยท `crypto_market` ยท `crypto_trending` ยท `crypto_movers` | $0.01 | Market data for trading agents |
| `fx_rate` ยท `weather` ยท `wiki_summary` | $0.01 | Currency, weather & Wikipedia utilities |
## ๐ก๏ธ One-line guard (make any agent safe to pay)
Wrap your x402 paying `fetch` โ DYOE verifies every payee **before** money moves, and blocks scams automatically:
```js
import { wrapFetchWithPayment } from "x402-fetch";
import { dyoeGuard } from "dyoe-agent-tools-mcp/guard";
const pay = wrapFetchWithPayment(fetch, account); // your normal x402 fetch
const safe = dyoeGuard(pay); // โ one line. every payee now verified.
await safe("https://some-seller.example/thing"); // throws automatically if it's a scam
```
For high-stakes actions, require a real human sign-off:
```js
import { requireApproval } from "dyoe-agent-tools-mcp/guard";
const decision = await requireApproval(pay, { action: "wire funds", counterparty: "acme.com", amount: 5000 });
// decision.decision === "approved" | "denied", with a signed, verifiable authorization
```
## Install
Add to your MCP client config (e.g. Claude Desktop / Cursor):
```json
{
"mcpServers": {
"dyoe-agent-tools": {
"command": "npx",
"args": ["-y", "dyoe-agent-tools-mcp"],
"env": {
"WALLET_PRIVATE_KEY": "0x<your-funded-base-burner-wallet-private-key>"
}
}
}
}
```
## ๐ณ Paying with x402
Each call spends USDC on **Base mainnet**, automatically, via x402 โ no API keys, no subscriptions.
- **`WALLET_PRIVATE_KEY`** โ a Base wallet that holds a little USDC. Payments are gasless for you (the facilitator covers gas).
- โ ๏ธ **Use a dedicated, low-balance burner wallet โ never your primary.** The key stays on **your** machine and signs payments locally; it never leaves your device and is never sent to any registry or to DYOE.
- **No wallet?** The server still runs in **free PREVIEW mode** โ it returns each tool's price quote instead of calling, so you can explore before funding a wallet.
## Links
- Live storefront & manifest: https://agents.dyoeway.org
- Built by [DYOE Way](https://dyoeway.org)
## License
MIT
TDQS
Scored across 16 tools
Tools have distinct purposes overall, with crypto, web extraction, trust, and general utilities clearly separated. Some overlap exists among trust tools (trust_check vs merchant_trust vs site_audit), but descriptions sufficiently differentiate them.
Naming uses snake_case but conventions vary: many tools have prefix patterns (crypto_, extract_) while others mix noun_verb (guardian_approve) or have inconsistent suffixes (intel_dossier vs token_intel). No strong uniform pattern.
16 tools is slightly above the typical well-scoped range but still acceptable. The broad scope covering crypto, web, trust, weather, wiki, etc. justifies the count, though it feels slightly heavy for a coherent server.
The tool set covers many common agent needs but lacks a focused domain. Gaps are unclear because it's a random collection; it doesn't provide complete CRUD for any specific area. Incompleteness is notable for a general toolkit missing obvious utilities like file handling or email.