1. Conduid
  2. Agents
  3. AgentLeak
MCP server · Agents

AgentLeak

AgentLeak (IEEE Access 2026): benchmark for internal-channel privacy leakage in multi-agent LLM systems. Inter-agent messages leak 68.8% vs 27.2% in final outputs; output-only audits miss 41.7%. 1,000 scenarios, 7 channels, 5 LLMs. Tool & MCP server: github.com/yagobski/agentleak

27Low

Scored 5 hours ago · breakdown

About AgentLeak

AgentLeak is an MCP server published by Privatris in the Agents category: agentLeak (IEEE Access 2026): benchmark for internal-channel privacy leakage in multi-agent LLM systems. Inter-agent messages leak 68.8% vs 27.2% in final outputs; output-only audits miss 41.7%. 1,000 scenarios, 7 channels, 5 LLMs. Tool & MCP server: github.com/yagobski/agentleak.

The repository has 36 stars and 4 forks, with the last commit a month ago. Recent commit activity is the strongest signal that reported issues will get attention.

Install

From source
git clone https://github.com/Privatris/AgentLeak

This server has no ConduID identity, so agent calls to it are not receipted. Pin the version you install and review the source before granting it credentials.

Ask AI

✦

Ask AI about AgentLeak

Powered by Claude

Ask about AgentLeak — answers draw on its README.

Security checks

6 checks have not been run yet.

Questions

About AgentLeak

How do I install AgentLeak?

No package has been published for this server, so install it from source: clone the repository and follow its README.

Is AgentLeak safe to use with an AI agent?

Its trust score is 27 out of 100 (low). Conduid hasn't run static security checks on this repository yet, so review the source yourself before granting it credentials. It has no ConduID identity yet, so agent calls to it are not receipted.

Is AgentLeak still maintained?

Yes — the last commit was a month ago. The repository has 36 stars and 4 forks.