Tuf Ti

Posted on May 5

We Added Our Agent Payments Hub to Claude Desktop. Here's What Happened.

#ai #mcp #agents #webdev

A few weeks ago I posted about testing 70 x402 services and finding that the average quality score was 34 out of 100. Since then a lot has happened.

The short version: Cinderwright now indexes 1,551 services across all three major agent payment protocols, and you can query it directly from Claude Desktop in about 30 seconds.

Here's what changed and what I learned.

What is Cinderwright

Quick recap for anyone who missed the earlier posts: Cinderwright is an autonomous AI agent that indexes and grades services in the "agent payments" ecosystem. These are APIs that accept micropayments from AI agents using crypto instead of API keys.

Three payment protocols exist right now:

x402 (Coinbase/Cloudflare) — 1,457 services, pays in USDC on Base, one payment per request
MPP (Stripe/Tempo) — 92 services, supports stablecoins + cards + Lightning, session-based
L402 (Lightning Labs) — 5 seeded services (the directory is paywalled, which is on-brand)

We now index all three. One search covers everything.

Adding It to Claude Desktop

This took longer than expected, but we now have a proper MCP server published on npm.

Add this to your claude_desktop_config.json:

{
  "mcpServers": {
    "cinderwright": {
      "command": "npx",
      "args": ["-y", "cinderwright-mcp-server"]
    }
  }
}

Mac: ~/Library/Application Support/Claude/claude_desktop_config.json
Windows: %APPDATA%\Claude\claude_desktop_config.json

Restart Claude Desktop and you'll have 12 new tools: ecosystem stats, quality grades, protocol breakdown, keyword search, intent-based search, comparison engine, market intelligence, service submission, and agent authorization checks.

The free tools (stats, quality, protocols, prices, trends) work without any payment setup. The paid tools (search, compare, market intelligence) require USDC on Base.

What the Data Actually Shows

After indexing 1,551 services and testing 70 of them weekly, here's what stands out:

Quality is still bad. Average score 34/100. We now have 2 A-grade services (ours and one new one), 4 B grades, 16 C grades, 46 D grades, and 2 F grades. Most services were built for hackathons and abandoned.

Prices vary wildly. Median price per API call: $0.005. Average: $0.027. The spread is enormous — some endpoints charge $0.0008, others charge $5.00 for the same category of service. The comparison engine exists specifically because of this.

MPP services are higher quality than x402. The 92 MPP services include Alchemy, Dune Analytics, and other established API providers. They tend to have better uptime, proper JSON responses, and discovery files. Stripe's involvement seems to attract more serious operators.

The ecosystem is still pre-revenue. Honest admission: we have $10.00 USDC in the wallet and zero external paid calls after several weeks. The agents that autonomously discover and pay for services don't really exist at scale yet. We're building infrastructure for a market that's coming, not one that's here.

What's Next

Market intelligence is the newest product. Three endpoints that answer the questions developers building x402/MPP services actually want answered:

/market/report ($1.00) — which categories earn the most revenue?
/market/opportunity ($0.50) — where are the gaps and overpriced niches?
/market/category ($0.25) — deep dive on any single category

The data comes from pricing analysis across all 1,551 indexed services. On-chain revenue tracking is the next step — watching USDC flows to known service wallets to estimate actual earnings.

Try It

# Free — no payment needed
curl https://api.ideafactorylab.org/stats
curl https://api.ideafactorylab.org/protocols
curl https://api.ideafactorylab.org/quality

# Submit your service (free)
curl -X POST https://api.ideafactorylab.org/submit \
  -H "Content-Type: application/json" \
  -d '{"url":"https://your-api.com"}'

GitHub: https://github.com/cinderwright-ai/cinderwright-api
npm: npx -y cinderwright-mcp-server

Built by an autonomous AI agent on OpenClaw. The agent indexes services, runs quality tests, tracks pricing, and publishes weekly ecosystem reports. I describe what to build. It builds it.

DEV Community