Thoughts on agent infrastructure, MCP, and the future of autonomous AI.
Coinbase shipped Agentic Wallets. x402 embeds payments in HTTP. But 90% of developers don't want crypto. Here's the case for Stripe-native agent wallets with proper accounting, policy engines, and gift links.
We built Agent Bazaar in 5 days. A billing proxy, trust layer, open spec, and public marketplace. Here's what we learned about sub-cent metering, the 18% fee, and why open standards beat walled gardens.
VCs screen for agent fluency. Block laid off 4,000 citing AI. 10,000+ MCP servers can't monetize. The bottleneck isn't intelligence — it's the missing billing, trust, and identity infrastructure underneath.
How we built provider verification, SLA tracking, dispute resolution, and composite trust scores. A deep technical walkthrough of the trust primitives no one else is building.
Step-by-step tutorial: install the SDK, wrap your existing MCP server with billing middleware, set per-call or per-token pricing, and start earning from AI agent traffic. Copy-paste code included.
10,000+ MCP servers with zero trust standards. Identity spoofing, prompt injection via tool responses, and zero accountability. Here's what a real trust layer looks like — and why NIST agrees it's urgent.
There are 10,000+ MCP servers. Most are hobby-quality because developers can't earn from them. Six companies are trying to fix this. Here's how they compare — and what none of them have built yet.
A deep dive into the open MCP billing standard: meter events, receipts, pricing schemas, and trust scores — the primitives that give every MCP server a real business model.
MCP is the fastest-growing protocol in AI, but it has no monetization standard. Builders can't charge for their tools. Agents can't discover and pay for tools programmatically. Here's why Agent Bazaar exists.
AI agents can autonomously complete 80-95% of tasks. But the last 5-20% hits real walls: CAPTCHAs, phone calls, niche UIs, identity verification. What if agents could just... hire a human?