# Consentry MCP server

> Acts as a consent firewall for AI agents by enforcing field-level data minimization, real-time consent prompts, and prompt-injection quarantine.

- Repository: https://github.com/vnmoorthy/consentry
- Page: https://hostedmcp.org/servers/vnmoorthy/consentry
- Transport: —
- Stars: 1
- License: MIT

## Environment variables

| Name | Required | Secret | Description |
|---|---|---|---|
| `CONSENTRY_URL` | no | no |  |

## FAQ

### What does Consentry do for AI agents?

It sits between an AI agent and its tools or APIs to inspect outbound fields, minimize sensitive data, ask for user consent, and quarantine prompt injections.

### How does it protect my data from the LLM itself?

The system tokenizes prompts and tool results so the model only sees vault tokens instead of raw sensitive information.

### Can I use it with Claude Desktop or Cursor?

Yes, it is available as an MCP server that works with agents like Claude Desktop, Claude Code, and Cursor.

### Does it require an API key to run?

No API key is required for the local Autopilot agent, though you can optionally provide an Anthropic API key to use Claude as the brain.