Top AI Product

DCP is a security layer for AI agents built on a simple inversion: agents get permission, not keys. Your Solana wallets and API credentials stay on your laptop. When an agent needs to act, it asks for permission, and you tap once on your phone to approve. Local-first, open source, works with Claude, Cursor, OpenClaw, and any MCP agent.

## The problem it targets

MCP has a credential-leak problem at scale — over 42,000 exposed MCP endpoints were found leaking credentials in 2026. The standard pattern hands an agent raw API keys and wallet secrets, so a compromised MCP server means an attacker walks away with everything. DCP adds an approval layer: even if the server is compromised, raw keys never leave your device.

## Why it matters

As agents gain the ability to spend money — crypto wallets, API billing — “give the agent your keys” stops being acceptable. DCP’s phone-approval model mirrors how hardware wallets fixed crypto custody: the secret never leaves secure hardware, you just authorize each action. Agent security is becoming a category, and key custody is the first battleground.