What is DCP?

DCP is an open-source, non-custodial security layer designed to manage AI agent permissions. By keeping API keys and digital wallets encrypted locally on your machine, it ensures that your AI agents only have the specific, scoped access you explicitly grant them.

Why Founders Need It

As you scale your operations using AI agents for real-world tasks, giving them broad access to company credentials or financial assets creates a significant security risk. DCP solves this by acting as a ‘permission gate,’ preventing unauthorized actions while maintaining high-velocity AI workflows.

How to Use It

  • Install the DCP environment locally.
  • Connect your AI models (e.g., Claude, Cursor) to the DCP layer.
  • Define granular scopes for each agent.
  • Approve agent actions in real-time via Telegram or the dedicated app.

Key Security Features

  • Local Encryption: Your secrets never touch the cloud.
  • Budget Controls: Set hard daily spending limits for agents.
  • Instant Revocation: Kill an agent’s access immediately if suspicious behavior is detected.
  • Audit Logs: Full transparency into what your agents are doing with your keys.

Integrations

Claude, Cursor, OpenClaw, Hermes, Model Context Protocol (MCP)