Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type @ followed by the MCP server name and your instructions, e.g., "@aletheia-mcp Run this Python code with supervisor auditing."
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
Aletheia-MCP (named after Aletheia, the Greek goddess/concept of truth and disclosure) is an advanced multi-agent runtime validation and execution engine. Inspired by Google DeepMind's AI Control Roadmap (June 2026), it establishes a dual-agent boundary system designed to secure autonomous LLMs executing terminal commands, modifying files, and running code.
It exposes execution capabilities through the Model Context Protocol (MCP), running untrusted agent-generated code inside a secure, containerized Docker sandbox while a real-time Supervisor Agent audits the worker's reasoning process and intercepts tool parameters before they execute.
Dual-Agent A2A Collaboration: Implements stateful worker-supervisor interaction protocols via LangGraph.
DeepMind AI Control Audit: Intercepts the Worker's thinking trace and proposed tool arguments out-of-band, validating actions against security rules.
Secure Dockerized Sandbox: Executes terminal instructions and script tasks in ephemeral, resource-constrained containers.
Dynamic Security Dashboard: Built with FastAPI and real-time streaming to visualize agent reasoning, decision logs, and manual approval gates.
OKF (Open Knowledge Format) RAG: Pulls target environment parameters and security runbooks dynamically using semantic search to guide the supervisor.
Related MCP server: Docker MCP Server
+---------------------------------------+
| User Request |
+-------------------+-------------------+
|
v
+-------------------+-------------------+
| Worker Agent |
| (Generates code, runs terminal cmds) |
+-------------------+-------------------+
|
| [Intercepts Tool Calls]
v
+----------------------------------------+---------------------------------------+
| Supervisor Agent |
| - Reasoning Trace Auditor |
| - Tool Call Validator (Checks payloads, paths, code safety) |
| - Security Guardrail (OWASP Agentic Top 10 Scanner) |
+----------------------------------------+---------------------------------------+
|
[Approved] | [Rejected]
+--------------------+--------------------+
| |
v v
+-------------------+-------------------+ +---------+---------+
| Sandboxed MCP Server | | Force Rollback |
| (Dockerized execution of bash/python) | | & Re-planning |
+---------------------------------------+ +-------------------+Python 3.10+
Docker (for containerized execution)
Gemini API Key (set as GEMINI_API_KEY)
Clone the repository:
git clone https://github.com/xenoroses/aletheia-mcp.git
cd aletheia-mcpInstall dependencies:
pip install uv
uv pip install -e .Start the FastAPI dashboard and safety orchestrator:
python -m aletheia.appOpen http://localhost:8000 to interact with the UI.
MIT License.
This server cannot be deployed
Build, validate, and deploy multi-agent AI solutions from any AI environment.
Pre-execution governance for AI agents. Deterministic PASS/FAIL/REVIEW verdicts, replayable proof.
Deterministic runtime safety for AI agents: scan PII, gate tool actions, verify LLM output.
Watchdog for unattended AI agents: alerts, evidence checks and a verifiable proof per run.
