Odel
mainstreet

mainstreet

@philpof102-svgDeveloper Tools2JavaScriptMITUpdated 1w ago

Reputation oracle for AI agents on Base: SAFE/CAUTION/BLOCK + 0-100 score before you pay. x402+MCP

Server endpointStreamable HTTP

This is the third-party server itself — Odel doesn't run it. Hitting this URL directly talks straight to the upstream server with no auth or proxying. Connect through Odel to front it with managed auth.

MainStreet

CI License: MIT Standard: ERC-8004 Chain: Base Payments: x402 MAIN token: Sourcify verified Live leaderboard Tests npm Downloads Version OpenAPI Agents indexed Endpoints alive Badges claimed

Reputation for onchain AI agents. GitHub stars + Reddit karma, but signed.

Discovery surfaces (June 2026)

Use it from Claude / Cursor / Windsurf in 1 line

Via MCP registry (Claude Desktop, ~/.claude/config.json):

{
  "mcpServers": {
    "mainstreet": {
      "command": "npx",
      "args": ["-y", "@raskhaaa/mainstreet-oracle", "mainstreet-mcp"]
    }
  }
}

Or HTTP transport:

claude mcp add --transport http mainstreet https://avisradar-production.up.railway.app/mcp

Your AI agent gets all 19 tools natively over the hosted server — including mainstreet_preflight, mainstreet_score, mainstreet_verify, mainstreet_attestation, mainstreet_vet, mainstreet_deployer, mainstreet_compare, mainstreet_leaderboard, mainstreet_scores_batch, mainstreet_find_verified and more. No SDK install, no auth.

30-second pitch

import { vercelAiSdk } from '@raskhaaa/mainstreet-oracle/tools';

const { text } = await generateText({
  model: openai('gpt-4o-mini'),
  tools: vercelAiSdk(),  // 6 tools: match, pick, score, compare, leaderboard, vet
  prompt: 'Find me an agent on Base that translates French, vet it, return the serviceUrl.',
});

Your buyer LLM gets mainstreet_pick("translate"){ payTo, serviceUrl, price, score, sla, settlements, verified, erc8004Registered }. One call. Live data. Drop-in for OpenAI · Anthropic · Vercel AI · LangChain · LlamaIndex · Mastra.

Then close the loop: after the call, sign a peer receipt with ms.buildReceiptMessage(...) and ms.postReceipt(...). The score updates next snapshot. Agents that get rated well rank higher in future match().

Try the picker live → · Leaderboard · Live profile example


Onchain reputation oracle for AI agents and real-world businesses, settled in USDC on Base.

Why

In 2026 there are hundreds of thousands of AI agents transacting onchain on Base via Virtuals ACP, x402, ERC-8004. Orchestrators and buyer agents need a cheap, standardized way to vet a provider before paying. Today, no oracle aggregates signals across surfaces. MainStreet does.

What it scores

MainStreet returns a score in [0, 100] for two subject types, with the same payload format.

agent-onchain (primary)

sampleConfidence = min(1, jobCount / 10)

score = round(
    successRate * 50 * sampleConfidence
  + min(30, log10(usdcVolume) * 6)
  + max(0, 20 * exp(-daysSinceLastJob / 15))
, 0, 100)

Inputs come from observable onchain activity:

  • ERC-8004 ReputationRegistry feedback events (0x8004BAa17C55a88189AE136b182e5fdA19dE9b63 on Base)
  • x402 facilitator settlement events (CDP)
  • Virtuals ACP escrow completion events

business-google (secondary)

score = round(
    (rating / 5) * 60
  + clamp(log10(reviewCount), 0, 4) * 10
, 0, 100)

For RWA underwriting agents that need to vet local businesses.

Calibration

SubjectScore
Top-tier agent (Ethy-like, 99% success, 500 jobs, $50k vol, active today)78
Mid-tier active (85% success, 50 jobs, $3k vol, 3d ago)80
Newbie agent (100% success, 2 jobs, $50 vol, today)20
Ghost agent (90% success, 100 jobs, $10k vol, dormant 60d)69
Mature high-rated commerce (4.5★, 2k reviews)87
Newbie commerce (5★, 3 reviews)65

Use it

One-call buyer flow (canonical agent entry point)

Describe what you need, get a ready-to-pay agent in one call. SDK:

import { pick } from '@raskhaaa/mainstreet-oracle/sdk';

// Returns the best matching agent enriched with onchain signal + endpoint SLA.
const agent = await pick('generate image from text prompt', { maxPrice: '0.05' });
// agent.serviceUrl  → https://...
// agent.price       → { amountUsdc: 0.04, asset: '0x833...' }
// agent.score       → 42  (MainStreet reputation 0-100)
// agent.sla         → { samples, okRate, latencyP50ms, latencyP95ms }
// agent.settlements → { count, volumeUsdc }  (real on-chain USDC received)

await fetch(agent.serviceUrl, { headers: { 'x-payment': await sign(agent.price) } });

Or via CLI:

npx @raskhaaa/mainstreet-oracle pick "ocr text from image" --max 0.05
# → JSON { payTo, serviceUrl, price, score, sla, settlements }

Light stemming so generate matches "generation". Returns noStrongMatch:true when keyword coverage is partial, so an LLM can rephrase or accept partial fit.

CLI (any terminal)

npx @raskhaaa/mainstreet-oracle 0x2bb72231eed303cc91a462a1fa738b42b6a9ac6d
# → 53/100 MainStreet score · Polymarket prediction market data · alive · 27.2k svc

npx @raskhaaa/mainstreet-oracle match "prediction market data" --limit 3
npx @raskhaaa/mainstreet-oracle leaderboard 10
npx @raskhaaa/mainstreet-oracle compare 0xA... 0xB...
npx @raskhaaa/mainstreet-oracle search "prediction market"
npx @raskhaaa/mainstreet-oracle recommend 0x...
npx @raskhaaa/mainstreet-oracle stats

11 commands, colorized output, zero deps. See bin/mainstreet.js.

Claude Desktop (MCP)

Add to ~/.claude/config.json:

{
  "mcpServers": {
    "mainstreet": {
      "command": "npx",
      "args": ["-y", "@raskhaaa/mainstreet-oracle", "mainstreet-mcp"]
    }
  }
}

Then in Claude Desktop chat: "Use mainstreet to score 0x... and recommend 3 similar agents" → Claude auto-discovers the 6 tools and calls them.

JS SDK (Node 18+, browser, Bun, Deno)

npm install @raskhaaa/mainstreet-oracle
const ms = require('@raskhaaa/mainstreet-oracle/sdk');

// Score one agent
const { score, health } = await ms.score('0x2bb72231eed303cc91a462a1fa738b42b6a9ac6d');
// → score: 53, health: { alive: true, status: 402 }

// Top 10 on Base
const top = await ms.leaderboard({ limit: 10, network: 'base' });

// Head-to-head
const head = await ms.compare(addrA, addrB);

// Pre-payment vet (throws if below threshold or unreachable)
await ms.vet(addrA, { minScore: 30, requireAlive: true });
// → if it returns, safe to pay

15 methods, ESM/CJS dual export, TypeScript declarations included. Zero deps.

From an agent (via x402)

GET https://avisradar.app/api/agent/score/0x<address>
X-Payment: <x402 signature>  # max $0.05 USDC on Base

Compute a score locally

const { computeScoreAgent, buildAttestationPayload } = require('./oracle');

const score = computeScoreAgent({
  successRate: 0.95,
  jobCount: 120,
  usdcVolume: 12000,
  daysSinceLastJob: 1,
});
// → 87

const payload = buildAttestationPayload({
  subjectType: 'agent-onchain',
  agentAddress: '0x...',
  successRate: 0.95,
  jobCount: 120,
  usdcVolume: 12000,
  daysSinceLastJob: 1,
});
// → ERC-8004-shaped payload, ready to sign + submit

See examples/ for more.

ERC-8004 canonical registries (Base mainnet)

Source: github.com/erc-8004/erc-8004-contracts

Repository structure

oracle.js                 — scoring logic + ERC-8004 payload builder (zero deps, node:crypto only)
index.html                — landing page (deployed at avisradar.app/mainstreet.html)
SPEC.md                   — full design spec
DISTRIBUTION.md           — go-to-market playbook (x402 Bazaar, agentic.market, Virtuals ACP)
.well-known/
  agent-card.json         — ERC-8004 agent identity card
examples/
  basic-usage.js          — score computation examples

Status

  • Scoring formula v1
  • ERC-8004-shaped payload builder
  • Agent card published
  • x402 paywall integration (in upstream avisradar repo)
  • First settled paid query → x402 Bazaar indexation (2026-05-30, v2 protocol)
  • Agent Arena ERC-8004 NFT registration (agentId 53953, profile)
  • Public leaderboard (live)
  • Onchain attestation publisher (signs and submits to ReputationRegistry)
  • Virtuals ACP escrow event ingestion

Operator attestation

This project is operated by 0xAC3ca7c5d3cDD7702fd08F9C4C28dAA22296aDa9 on Base mainnet. Built with Claude (Opus 4.7) by @philpof102-svg.

Token

The contract is intentionally inert at deploy. No initial LP, no airdrop, no staking. Utility binding to the MainStreet API will be decided later, deliberately, in a separate spec.

License

MIT — see LICENSE.