About provenance-mcp
provenance-mcp checks Algorand x402 endpoints before an agent pays, using public on-chain payments to estimate whether revenue may be manufactured through wash activity or self-dealing. Its free tool returns a risk level, score, and signals; scores are estimates, not judgments about endpoint operators.
Key features
- 01Check Algorand payee address risk
- 02Return risk level and 0–100 score
- 03Show payer, wallet, and funding signals
- 04Offer paid, in-depth forensic endpoints
Use cases
- 01Screen an x402 API before payment
- 02Use on-chain signals to inform payment decisions
MCP support
Exposes endpoint risk checks and service information to MCP clients. The risk tool analyzes Algorand on-chain payments and returns an estimate with supporting signals.
Configuration
Local server config is the same for every client. Copy it into your client’s MCP config file.
We haven’t collected a package or remote endpoint for this server. Follow the instructions in the project README.
Tools
2Saved tool list · updated 2w ago
| Parameter | Type | Required | Description |
|---|---|---|---|
| address | string | Yes | The Algorand endpoint payTo address (58 characters, A-Z and 2-7) |
Details
From the project READMEprovenance-mcp
Check the wash-traffic risk of Algorand x402 endpoints before your agent pays them.
📊 The x402 Trust Index — our free public data site: dated Bazaar Wash Report snapshots (Base) + reseller-signal audits for 38 API brands.
Leaderboards for machine-payable APIs rank by claimed volume, which is easy to inflate — self-dealing loops, fresh-wallet farms, metronomic bots. Provenance reads public on-chain payments and gives your agent a statistical estimate of whether an endpoint's revenue looks organic or manufactured, so it can decide before sending USDC. This server scores Algorand payees; the public data site's report covers Base. Scores are one operator's estimates — the methodology publishes the weights, the limits, and what has not been validated yet.
Tools
check_endpoint_risk(address)— free quick verdict for any Algorand endpoint payTo address: risk level (low/medium/high/critical), 0–100 score, and the on-chain red flags (distinct-payer count, self-dealing payout loops, wallet ages, timing regularity, funding concentration).provenance_service_info()— service + paid-tier info.
Deep forensics are x402-paid on the same API: /score $0.05 · /report $0.50
(full 8-signal grade + facilitator drift) · /diligence $5.00.
Install
Claude Code
claude mcp add provenance -- npx -y provenance-mcp
Claude Desktop / any MCP client (mcpServers config):
{
"mcpServers": {
"provenance": {
"command": "npx",
"args": ["-y", "provenance-mcp"]
}
}
}
Then ask your agent: “check the wash risk of Algorand endpoint <ADDRESS>”.
How it works
The hosted Provenance engine indexes USDC (ASA 31566704) payments on Algorand mainnet, enriches payer wallets (age, first funder, asset diversity, rekey auth-addr), clusters them (union-find over funding + shared-key links), and runs 8 wash-detection signals. Scores are point-in-time snapshots, anchored on-chain for tamper-evidence. Methodology: signals, not accusations — a high score means the revenue pattern is consistent with self-dealing, not that the operator is guilty of anything.
- Env:
PROVENANCE_API_URLto point at a different Provenance deployment. - Free tier: 30 checks/hour/IP.
Guard your x402 client directly
Building an agent that pays over x402? provenance-guard
wraps any x402-paying fetch and runs this same check automatically before any
money moves — one line, zero dependencies:
import { createGuard } from "provenance-guard";
const guarded = createGuard().guardedFetch(fetchWithPayment);
MIT · Provenance — the trust layer for machine-payable endpoints.
Comments
New to MCP?
From what the protocol is to installing your first server, these answers cover most first questions.
An open protocol that defines how AI applications connect to external tools and data in a uniform way. Clients talk to the model; MCP servers provide tools, resources and prompt templates. They exchange standard messages, so one server can be reused by many clients.