OP here. I built this because I realized my AI agents were 'anonymous ghosts' with no way to prove their identity to APIs.
If an agent hallucinates or gets hijacked, there's no audit trail. This is a simple MCP server that gives agents a local RSA wallet so they can cryptographically sign their own actions.
It's open source and works with Claude Desktop out of the box. Would love feedback on the security model—specifically if I should move to Ed25519 keys for the next version.
faalantir•2mo ago
If an agent hallucinates or gets hijacked, there's no audit trail. This is a simple MCP server that gives agents a local RSA wallet so they can cryptographically sign their own actions.
It's open source and works with Claude Desktop out of the box. Would love feedback on the security model—specifically if I should move to Ed25519 keys for the next version.
Repo: https://github.com/faalantir/mcp-agent-identity