Interactive Tool

MCP Security Policy & Config Generator

Configure your AI coding agent with hardened, production-safe MCP server settings. Select your client, target service, and safety profile to generate copy-paste configuration files and deterministic gate rules.

Zero external network requests · 100% self-hosted · Safe configuration in under 30 seconds

1. Select your agent client

2. Select target service or connector

3. Select security policy level

Strict Read-Only Agent can only query data (SELECT). All DROP, ALTER, TRUNCATE, INSERT, and UPDATE statements are blocked deterministically.

Protected Write (Recommended) Allows safe single-row writes; pauses bulk updates (> 5 rows) and schema alterations for mandatory human confirmation.

Full Read-Back & Audit Verifies committed physical state in the database WAL change log after every tool call and seals tamper-evident Merkle deeds.

-95.2% Blast Radius Reduction

Grade A Security Rating

< 0.9 ms Local Gate Latency

Loading...

Save this to ~/.cursor/mcp.json

Why standard MCP servers need a deterministic gate

The MCP Execution Risk

Standard Model Context Protocol servers run tool calls with direct access to your underlying infrastructure. When an AI model hallucinates a table name, misplaces a WHERE clause, or encounters prompt injection, the tool executes unconditionally.

The Deterministic Defense

Deedglass runs as a local loopback gateway between your client and MCP servers. Policies check arguments before execution, physical database WAL logs verify state after execution, and every action is sealed in an offline-verifiable audit deed.