MirginCipher Blackbox (MGC) — Multi-Agent Cross-Node Privacy Execution Base

A local encrypted execution base for multi-agent, cross-node, privacy-preserving task execution — sensitive operations never leave the device in plaintext; scripts, files, and skill / workflow folders can be sealed and dispatched across trusted nodes without exposing source code.

📌 Roadmap: MGC Blackbox — 2026 Development Plan

What is MGC Blackbox?

MirginCipher Blackbox (MGC) is a Multi-Agent Cross-Node Privacy Execution Base. It is not an agent itself — it is the encrypted execution layer underneath your agents.

MGC provides:

• A local encrypted execution base for AI agents, system scripts, and human users
• Cross-node task delegation via sealed, node-bound script capsules
• Zero plaintext leakage: secrets stay inside the encrypted boundary
• A trusted device-level boundary shared by Copilot / Claude / Trae / IDE Agents

Why MGC?

• 🔐 End‑to‑End Encrypted Storage
  AES‑256 encrypted vault for API keys, credentials, configs — never exposed to AI agents or external systems.

• 🧱 Local‑First Security Boundary
  All execution and decryption happen on‑device. No cloud dependency, no plaintext leakage, no telemetry.

• 🧩 Sealed Script & Folder Execution (Unique)
  Convert scripts, files, and skill / workflow folders into unreadable, node-bound execution capsules.
  Only the target node can decrypt & run them — neither sender nor third parties can read the source.

• ⚡ Deterministic Local Execution
  Stable, reproducible behavior across macOS / Linux / Windows with a Cython‑compiled secure core.

• 🛠️ Native MCP / Skill Integration
  Exposes mgc_save / mgc_save_file / mgc_get / mgc_list / mgc_run / mgc_find / mgc_package / mgc_seal_package / mgc_open_webui / mgc_skill as standard MCP tools.
  Works out‑of‑the‑box with Copilot, Claude, Trae, IDE Agents.

• 🔄 Zero Integration Cost
  Any MCP‑compatible agent can immediately use MGC as its secure execution backend — no SDK, no custom code.

• 🛡️ Designed for AI Agent Security
  Protects human intent, prevents agent overreach, and enforces strict execution boundaries.

Use Cases

1. Protect API Keys & Credentials from AI Agents

Store secrets encrypted. Agents can use them, but never see plaintext.

2. Secure Local Automation

Run Python / Shell / Node scripts locally without exposing sensitive data to AI logs or cloud systems.

3. Sealed Script Distribution

Share scripts with collaborators or devices without exposing source code — they can execute but cannot read.

4. Cross‑Node Execution

Send sealed scripts to trusted remote nodes:

• Sender cannot read the sealed content
• Recipient cannot read the sealed content
• Only the target node can decrypt and execute

5. Local‑First AI Agent Security Boundary

Provides a local security layer for Copilot / Claude / Trae / IDE Agents.

6. Privacy‑Preserving AI Workflows

Enables financial automation, personal data processing, and enterprise internal workflows with privacy protection.

7. Skill / Workflow Folder Sealing

Bulk-import a skill / workflow folder via mgc_save_file(path=<local folder path>), then ship it as a sealed .mgc_file via mgc_seal_package — the recipient node decrypts and runs each script in its own encrypted boundary, with no source code exposure. (1.5.1+) MGC also auto-recognizes the cross-script call graph inside the folder (imports, from-imports, relative-path subprocesses) and routes those calls through sealed execution — no rewrites, no plaintext on disk.

📘 Value Scenarios

MGC Blackbox provides a trusted, encrypted execution boundary for different roles and environments.
Detailed scenario documents:

• 🔐 Sensitive Credentials Authorization
  docs/Sensitive_Credentials_Authorization.md

• 🧠 Encrypted Cognitive Script Execution
  docs/Encrypted_Cognitive_Script_Execution.md

• 🌐 Cross‑Node Execution Grant (Encrypted Capability Sharing)
  docs/Cross‑Node_Execution_Grant%20(Encrypted_Capability_Sharing).md

Architecture

Crypto Layer & Performance

MGC uses a hybrid cryptographic design:

• AES‑256‑GCM — bulk data encryption
• RSA‑2048/4096 — key encapsulation & node authorization

The crypto layer is Cython‑compiled to:

• Improve AES & RSA performance
• Reduce Python overhead
• Provide a sealed, tamper‑resistant execution boundary
• Maintain deterministic behavior across nodes

Features

• Local encrypted storage
• Encrypted execution
• Store‑once authorization
• Environment migration
• Cross‑agent availability
• Cross‑platform support
• No delete function (manual DB deletion only)
• Script sealing for cross‑node execution
• Cross-script auto-recognition in skill / workflow folders (1.5.1+) — mgc_save_file parses the script call graph (import / from-import / relative-path subprocess) at save time; mgc_run then routes cross-script calls through sealed execution transparently.

Quick Start

1. Install

pip install mgc-blackbox

2. Start Service

mgc

3. Open WebUI

http://127.0.0.1:57218

4. Store a Secret

from mgc import save
save("openai_key", "sk-xxxx")

5. Execute Scripts Securely

Scripts run inside MGC's encrypted boundary.

Example: Save & Retrieve Secrets

from mgc import save, get

save("openai_key", "sk-xxxx")
print(get("openai_key"))

MCP Integration

MGC exposes a local MCP tools interface:

• mgc_save — store a single secret / script / file by explicit content
• mgc_save_file — store by local file path; auto-detects single file vs folder; (1.5.1+) also auto-recognizes cross-script calls inside folder saves (import / from-import / relative-path subprocess, including pure-literal lists like subprocess.run(['helper.py']))
• mgc_get — retrieve encrypted content
• mgc_list — browse stored entries
• mgc_run — execute a stored script in the encrypted boundary; (1.5.1+) cross-script calls inside a folder run transparently as sealed child processes
• mgc_find — fuzzy-search entries by info_owner / diff_1..3 (no info_type)
• mgc_package — bundle a folder into a sealed .mgc_file
• mgc_seal_package — ship a sealed package to a trusted node
• mgc_open_webui — open the local WebUI
• mgc_skill — return the bundled skill document (skill_spec.md) for the agent

Compatible with Copilot, Claude, Trae, IDE Agents.

Usage Overview

1. Through AI agents (Skills / MCP)

Agents can store secrets, retrieve encrypted items, execute scripts, and seal scripts.

2. Through system scripts (REST API)

External scripts can fetch encrypted items at runtime.

Security Model

• All data remains local
• No cloud upload
• No plaintext logging
• Deterministic execution
• User‑controlled authorization
• Sandbox mode support (fallback hardware fingerprint for restricted environments)

AI Skill Specification

See: docs/skill_spec.md

Authorization

Integration into any third‑party products or AI agents is free,
but requires official authorization to ensure ecosystem integrity.

Contact: mirgincipher@outlook.com

License

See the LICENSE file for full terms.

© 2026 MirginCipher Team. All rights reserved.

---