# Kadent > Kadent is a security gateway for MCP (Model Context Protocol). It sits between > AI agents — Claude, Codex, Cursor, or any MCP client — and the tool servers > they call. It records every tool call to an append-only, hash-chained, > Ed25519-anchored audit log, evaluates policy in the request path (allow, hold, > or deny), blocks lethal-trifecta data exfiltration, and produces evidence that > can be verified independently of Kadent itself. Key facts: - One MCP endpoint (mcp.kadent.so) replaces direct connections to every tool server; existing servers are added behind it unchanged. - The free scanner at https://kadent.so/scan performs a read-only MCP handshake against any server you are entitled to connect to and reports its tools, annotations, and risk posture. - Verification of the audit chain does not depend on Kadent existing: exports can be checked with a standalone script against the published signing key. - Free tier: 50,000 recorded calls/month, unlimited connected servers. Paid plans from $39/month. - No model ever makes an allow/deny decision; policy evaluation is deterministic and the audit record shows the rule that fired. ## Docs - [Documentation](https://kadent.so/docs): quickstart, policy reference, audit record schema, security model, data handling - [What MCP is and why a gateway](https://kadent.so/docs#primer): plain-language primer - [Connect your agent](https://kadent.so/connect): per-client setup for Claude Code, Claude Desktop, Cursor, VS Code, and any MCP client ## Product - [MCP server scanner](https://kadent.so/scan): free, no account required - [Pricing](https://kadent.so/pricing): metered on calls recorded, servers never limited - [Console](https://app.kadent.so): sign in with GitHub or an emailed link ## Legal - [Terms of service](https://kadent.so/terms) - [Privacy](https://kadent.so/privacy) - [Data processing terms](https://kadent.so/dpa) - [Sub-processors](https://kadent.so/sub-processors) Contact: team@kadent.so