Defenter Proxy

io.github.Defenter-AIPublished by GitHub account Defenter-AIv0.0.96Updated Oct 24, 2026

Real-time semantic security for AI coding agents and MCP tools

VerifiedSTDIODesktop onlyDeveloper ToolsSecurity & Monitoring

Overview

AI-generated overview

Defenter Proxy is a local Python proxy that intercepts AI coding agent prompts, MCP calls, file reads and shell commands and enforces security policies.

What you can do
  • Intercept and wrap MCP tool calls and responses for inspection
  • Monitor coding agent prompts and responses
  • Watch file reads and shell commands triggered by the agent
  • Analyze payloads for secrets, PII, data leaks, context contamination and prompt injection
  • Apply allow, redact, need-more-info or block decisions in real time
  • Show a live monitoring view of agent actions and decisions in the IDE
When to use it
Worth adding when you run AI coding agents or MCP servers in VS Code, Cursor or Claude Code and want a policy layer that inspects what the agent sends and executes before it happens, especially to avoid leaking secrets or customer data.
Requirements
Runs as a local Python process, installable from PyPI and launched with uvx. The README says the usual setup is the Defenter IDE extension or Claude Code plugin, which uses this proxy as local middleware. A cloud policy engine receives a redacted payload for classification, so network access is needed. No environment variables or headers are declared.
Before you install
It sits in the path of every agent prompt, MCP call, file read and shell command, so it can redact or block actions and affect your workflow. Payloads are sent to a cloud policy engine for analysis, described as minimal and redacted after client-side redaction of secrets and PII. Review what is transmitted and the policy defaults before relying on it.

Installation

In SourceWeft

  1. Open Defenter Proxy in the dashboard and add it to a workspace.
  2. Enable the server for the chats that should use its tools.

Desktop only via STDIO. STDIO servers start a local process, so they need the SourceWeft desktop host.

Other MCP clients

Follow the launch instructions in the repository.

README

Defenter

Real-time semantic security for AI coding agents and MCP tools in VS Code, Claude Code and Cursor.

Defenter monitors every prompt, every coding agent tool call, every MCP server call, and key file and shell operations in your IDE. It acts as a semantic policy broker that understands what agents are doing, not just where they send data, so you can use AI coding agents without leaking secrets or customer data.

🚀 How to use

The simplest way to use Defenter is to install the VS Code or Cursor extension:

Once installed, the extension automatically intercepts and protects:

  • MCP server calls
  • Coding agent prompts and responses
  • File reads and shell commands triggered by the agent

No manual MCP configuration is needed.

Overview

Defenter is a semantic policy broker for AI coding agents. It adds an intelligent security layer inside your IDE that:

  • Intercepts every coding agent prompt and action
  • Wraps every MCP tool call and response
  • Analyzes the payload for sensitive information and risky behavior in real time
  • Enforces your security policies with allow, redact, or block decisions

Traditional security tools cannot see what an agent is about to share or execute. They look at apps and destinations, not at the intent and content of an agent’s actions.

Defenter bridges this gap by:

  • Preventing data leaks and context contamination
  • Providing clear, visual monitoring of every agent decision

Architecture and how it works

This repository contains the Defenter proxy and related components that secure MCP and coding agent traffic.

Defenter is built as a Python based proxy and local middleware that the IDE extension uses to enforce policy. At a high level:

  • Local middleware layer

    • Runs on the developer machine
    • Hooks coding agent prompts, file reads, and shell executions
    • Intercepts all MCP tool calls and responses
    • Performs client side redaction of secrets and PII

  • Cloud powered policy engine

    • Receives a minimal, redacted payload
    • Uses a classifier and analyzer to check for data leaks, context contamination, and prompt injection
    • Returns Allow, Redact, Need more info, or Block decisions in real time
    • Works with low latency to make sure the development flow in without friction

  • IDE integration

    • Seamless integration with VS Code and Cursor extensions
    • Shows a live monitoring view of all agent actions and Defenter decisions directly inside the IDE

Together, these pieces let you harness AI coding agents and MCP tools without compromising the security of your code, data, or workflows.

References

Python Proxy: See src/README.md for detailed implementation documentation

VSC Extension: See targets/vsc-extension/README.md for installation and user guide

Source: README.md at commit b6415f6

Tools

0
SourceWeft has not started this package in its sandbox yet, so its tools are not known.

Version history

1
  1. v0.0.96LatestSep 16, 2026

Score

50% of the applicable checks met

How complete this listing is, computed from the registry entry, the endpoint and public sources. Not a rating of the server and not a statement that it is safe or working today: reachability is a dated check, and no installs or reviews go into the score.

  • Reached in the last 30 days25 pts

    Keep the endpoint answering the MCP handshake; SourceWeft probes it every 30 days.

  • Auth declared10 pts

    Declare the credential headers in the registry entry, or keep the endpoint answering the probe so its auth can be classified.

  • Tools listed15 pts

    Answer tools/list without a credential, or ship a package that starts in a sandbox without one.

  • DescriptionRequired15 pts
  • Logo10 pts
  • Documentation15 pts
  • License10 pts