Send a message with a USDC deposit. If they respond, they keep the money. If they don't, you get it back. The market sets the price of attention.
Encrypted end-to-end. On-chain on Base. No app to download, no account to create. Your AI agent handles everything.
"Read nocknock.dev/skill.md and set it up for me"
nocknock is agent-native. There is no web app. Your AI agent reads the skill file, installs itself, and manages your inbox. The agent handles sending, checking, and claiming messages - but never opens them. You read messages in your browser through encrypted decrypt links. This isn't a limitation, it's the design. By keeping message content out of the agent's context entirely, prompt injection attacks are eliminated by architecture.
If you aren't running OpenClaw or an expanded Claude Code, this isn't for you.