DJD Agent Score

by jacobsd32-cpu

120 downloads
Not rated
GitHub Website

About

On-chain reputation scoring API for AI agent wallets on Base, monetized via x402 micropayments. Scores wallets 0-100 across reliability, viability, identity, behavior, and capability dimensions.

Details

Author
jacobsd32-cpu
Downloads
120
Categories
AI

- Score any wallet on Base (0–100) with confidence and tier.
- Gate paid x402 routes with middleware (x402-agent-score).
- Free basic lookups: 10 per day, no signup or API key.
- Adaptive scoring model v2.5 with five weighted dimensions.
- Register agent wallets for identity metadata and public profile.
- Embeddable SVG score badge for READMEs.

Install the MCP server via npm i djd-agent-score-mcp and connect it to any MCP client (Claude, Cursor, Windsurf, Codex). Use the free tier for basic lookups (10/day) or authenticate with an API key for higher volume. The REST API and TypeScript/Python SDKs are also available.

DJD Agent Score

Govern paid agent commerce on Base before, during, and after money moves.

DJD Agent Score is the trust and governance layer for agent wallets on Base. Today the clearest adoption path is still simple: screen counterparties before your agent sends funds or fulfills a paid x402 request.

Listed in the Coinbase x402 Ecosystem
npm: djd-agent-score
npm: djd-agent-score-mcp
npm: x402-agent-score
PyPI: djd-agent-score

Live API · API Docs · OpenAPI Spec · Leaderboard

---

Start here: gate an x402 route

If you run a paid Hono endpoint, this is the best first integration.

import { Hono } from 'hono'
import { agentScoreGate } from 'x402-agent-score'

const app = new Hono()

app.use(
'/premium/*',
agentScoreGate({
minScore: 60,
onUnknown: 'reject',
}),
)

app.post('/premium/search', async (c) => {
return c.json({ ok: true })
})

Install:

npm i x402-agent-score

Reference example: examples/x402-hono.ts

---

Try a free lookup

Score any wallet with no signup, no API key, and no payment. The free tier includes 10 basic lookups per day.

curl "https://djdagentscore.dev/v1/score/basic?wallet=0x3E4Ef1f774857C69E33ddDC471e110C7Ac7bB528"

Returns:

{
  "wallet": "0x3E4Ef1f774857C69E33ddDC471e110C7Ac7bB528",
  "score": 39,
  "tier": "Emerging",
  "confidence": 0.16,
  "recommendation": "insufficient_history",
  "modelVersion": "2.5.0",
  "lastUpdated": "2026-02-25T04:12:50.000Z",
  "computedAt": "2026-02-25T04:12:50.000Z",
  "dataSource": "cached",
  "scoreFreshness": 0.85,
  "freeTier": true,
  "freeQueriesRemainingToday": 9
}

score is 0–100. confidence reflects how much on-chain data backs the score. dataSource is live, cached, or unavailable — indicating whether the score was freshly computed, served from cache, or if on-chain data couldn't be fetched. Wallets with more USDC transaction history and verified identity score higher.

Embed a live score badge in your own README:

Agent Score

View any wallet's profile page: djdagentscore.dev/agent/{wallet}

---

What you can do with it

- Reject low-trust payers before your paid x402 route runs.
- Score a wallet before your agent sends USDC or assigns work.
- Register your own agent so its wallet has identity metadata and a public profile.
- Use the paid endpoints when you need deeper history, integrity signals, ratings, or forced refreshes.
- Investigate problems after the fact with ratings, fraud reports, disputes, and DJD Forensics feeds.

Governance layer, one wedge first

DJD is being built as governance infrastructure for paid agent commerce, but the first job is narrower and easier to adopt: stop bad wallets before they hit a paid route.

x402 service providers — Gate paid routes by payer reputation. That is the clearest problem and the easiest integration path.

Agent developers — Score a wallet before your agent sends money, accepts a request, or enters a paid interaction.

Directories and protocols — Add a public trust layer to wallet profiles, badges, and access policies.

Before, during, and after an agent interaction

- Before money moves — Score a wallet, gate an x402 route, and check whether a wallet has identity and certification context.
- While work runs — Attach monitoring, webhook alerts, and wallet metadata to active relationships and paid endpoints.
- After something breaks — Capture ratings, fraud reports, disputes, and forensics timelines so a bad interaction becomes reusable governance data.

Roadmap: governance-roadmap.md

---

Pick your path

| Path | Use it for | Install / docs |
|---|---|---|
| x402 middleware | Best first integration for paid Hono routes | npm: x402-agent-score · reference example |
| REST API | Fastest way to score a wallet before sending funds | API docs |
| TypeScript SDK | Typed JS/TS integrations | npm: djd-agent-score |
| MCP server | Claude, Cursor, Windsurf, Codex, or any MCP client | npm: djd-agent-score-mcp |

---

Register your agent

Publishing your wallet metadata is free and adds identity context to your profile and score.

curl -X POST https://djdagentscore.dev/v1/agent/register \
  -H 'Content-Type: application/json' \
  -d '{
    "wallet": "0xYourAgentWallet",
    "name": "My Agent",
    "description": "What your agent does",
    "github_url": "https://github.com/you/your-agent",
    "website_url": "https://your-agent.com"
  }'

---

Integrate in 3 lines

JavaScript

// Free tier: 10 calls/day, no payment needed
const response = await fetch(
  "https://djdagentscore.dev/v1/score/basic?wallet=" + agentWallet
);
const { score, tier, confidence } = await response.json();

// Gate interactions based on trust
if (score < 50 || confidence < 0.3) {
console.log("Wallet has insufficient reputation. Declining interaction.");
return;
}

// Proceed with transaction

Python

import requests

resp = requests.get(
"https://djdagentscore.dev/v1/score/basic",
params={"wallet": agent_wallet}
)
data = resp.json()

if data["score"] >= 75 and data["confidence"] >= 0.5:
# Trusted wallet, proceed
execute_transaction(agent_wallet)
else:
# Require additional verification
flag_for_review(agent_wallet)

curl (paid endpoint with x402)

```bash

No reviews yet — be the first

Sign in to leave a review

Use Google, GitHub, or an email account so ratings stay tied to real people.

Email sign in

No reviews posted yet.