MCP server with zkDNA identity-based authentication for AI agent tool calls.
Add to your claude_desktop_config.json:
{
"mcpServers": {
"zkdna": {
"command": "npx",
"args": ["tsx", "/path/to/zkDNA/packages/mcp-server/src/index.ts"]
}
}
}npx tsx packages/mcp-server/src/index.ts| Tool | Required Capability | Description |
|---|---|---|
web_search |
search:web |
Search the web for information |
read_document |
read:documents |
Read a document by ID |
write_report |
write:reports |
Write a report |
agent_info |
(none) | View agent identity, capabilities, and audit log |
- Server starts with a principal DID (human identity)
- A default agent is registered with scoped capabilities
- Each tool call is authenticated (verify agent credential) and authorized (check capability policy)
- All actions are logged in a verifiable audit trail
- Unauthorized tool calls are blocked with clear error messages
zkdna://agent/identity— Current agent's DID, capabilities, and credential info