secure-mcp

Name	secure-mcp JSON
Version	0.1.0 JSON
	download
home_page	https://github.com/re-rank/secure-mcp
Summary	Secure execution layer for LLM/MCP agents
upload_time	2025-07-21 23:05:23
maintainer	None
docs_url	None
author	re-rank
requires_python	>=3.7
license	Apache-2.0
keywords	security mcp llm agent sandbox whitelist audit
VCS
bugtrack_url
requirements	No requirements were recorded.
Travis-CI	No Travis.
coveralls test coverage	No coveralls.

            # 🛡️ Secure-MCP

[![English](https://img.shields.io/badge/lang-English-blue.svg)](README.md)
[![한국어](https://img.shields.io/badge/lang-한국어-green.svg)](README-ko.md)
[![日本語](https://img.shields.io/badge/lang-日本語-red.svg)](README-ja.md)
[![中文](https://img.shields.io/badge/lang-中文-orange.svg)](README-zh.md)
[![Español](https://img.shields.io/badge/lang-Español-yellow.svg)](README-es.md)

> **Secure execution layer for LLM/MCP agents** — safely integrate agents without exposing `.env`, secrets, or sensitive system files.

---

## 🔍 What is Secure-MCP?

**Secure-MCP** is a security-focused middleware framework that wraps around your MCP (Multi-Component Prompting) or LLM Agent systems to ensure that:

- 🚫 `.env`, `API keys`, system files, and private configs are **not accessible** by agent logic.
- ✅ MCP/agent systems can still **safely perform operations**, such as calling APIs or retrieving tokens, via a secure proxy.
- 🧠 It supports LLM workflows, agent frameworks (like LangChain, AutoGen, CrewAI), and context-based tools — **without compromising security**.

---

## 🧱 Key Features

| Feature                         | Description                                                                 |
|--------------------------------|-----------------------------------------------------------------------------|
| 🔐 **SecureEnvBridge**          | Only whitelisted env variables can be accessed; `.env` is not directly exposed |
| 📂 **File Access Control**       | Blocks file reads like `.env`, `secret.pem`, `key.json`, `config.yaml` etc. |
| 📦 **Secure Function Proxy**     | Sensitive operations are abstracted into secured callable functions         |
| 📜 **Audit Logging**            | Logs all attempts to access sensitive resources                             |
| 🧪 **Sandboxed Execution**       | Optionally runs MCP code in an isolated environment (e.g., Docker, WASM)     |
| ⚙️ **Framework Agnostic**         | Easy to plug into any agent system or LLM pipeline                          |

---

## 🧰 Example Use Case

```python
from secure_mcp.secure_env_bridge import secure_get, secure_open

# Only whitelisted environment variables are allowed
api_key = secure_get("SERVICE_AUTH_TOKEN")  # ✅ allowed
db_pw = secure_get("DB_PASSWORD")           # ❌ blocked

# Safe file open (blocks sensitive files)
with secure_open("safe_data.csv") as f:
    content = f.read()
```

Use `SecureEnvBridge` instead of `os.getenv()` and `open()` to ensure safety.

---

## 🔧 Installation

```bash
git clone https://github.com/yourname/secure-mcp.git
cd secure-mcp
pip install -e .
```

You may also package it via `setup.py` or `pyproject.toml` for internal PyPI deployment.

---

## 🛠️ Architecture

```txt
[ User Input / Agent Task ]
           ↓
[ SecureContextExecutor ]
           ↓
[ SecureEnvBridge / FileGuard ]
           ↓
[ External API, File, or Memory Access ]
           ↓
[ LLM Agent Response ]
```

* MCP doesn't need to know the secrets.
* SecureEnvBridge acts like a **vault-aware delegate**.
* Every access is **controlled, logged, and filtered**.

---

## 📄 Example Structure

```
secure-mcp/
├── secure_env_bridge/
│   ├── secure_get.py
│   ├── secure_open.py
│   ├── whitelist_loader.py
│   └── audit_logger.py
├── examples/
│   ├── fastapi_demo.py
│   └── sandbox_runner.py
├── tests/
│   └── test_env_guard.py
├── .env.example
├── LICENSE
├── README.md
└── setup.py
```

---

## 🚧 Roadmap

* [x] SecureEnvBridge with whitelist loader
* [x] File access guard (`open()` patch)
* [ ] WASM-based execution wrapper
* [ ] Dockerized secure MCP runner
* [ ] OpenAPI proxy adapter
* [ ] Agent plugin system for LangChain / CrewAI

---

## 🔐 License

This project is licensed under the **Apache License 2.0**.
See the [LICENSE](./LICENSE) file for more details.

---

## 🤝 Contributing

Pull requests are welcome.
For major changes, please open an issue first to discuss what you would like to change.
Check `CONTRIBUTING.md` for detailed instructions.

---

## 📬 Contact

Questions, ideas, or feedback?
Feel free to [open an issue](https://github.com/yourname/secure-mcp/issues) or start a discussion.

Raw data

            {
    "_id": null,
    "home_page": "https://github.com/re-rank/secure-mcp",
    "name": "secure-mcp",
    "maintainer": null,
    "docs_url": null,
    "requires_python": ">=3.7",
    "maintainer_email": "Your Name <your.email@example.com>",
    "keywords": "security, mcp, llm, agent, sandbox, whitelist, audit",
    "author": "re-rank",
    "author_email": "Your Name <your.email@example.com>",
    "download_url": "https://files.pythonhosted.org/packages/9e/01/24eaca6c50cb898993152350d643ff5164ecca8e9579971ea864cf8a84fe/secure_mcp-0.1.0.tar.gz",
    "platform": null,
    "description": "# \ud83d\udee1\ufe0f Secure-MCP\r\n\r\n[![English](https://img.shields.io/badge/lang-English-blue.svg)](README.md)\r\n[![\ud55c\uad6d\uc5b4](https://img.shields.io/badge/lang-\ud55c\uad6d\uc5b4-green.svg)](README-ko.md)\r\n[![\u65e5\u672c\u8a9e](https://img.shields.io/badge/lang-\u65e5\u672c\u8a9e-red.svg)](README-ja.md)\r\n[![\u4e2d\u6587](https://img.shields.io/badge/lang-\u4e2d\u6587-orange.svg)](README-zh.md)\r\n[![Espa\u00f1ol](https://img.shields.io/badge/lang-Espa\u00f1ol-yellow.svg)](README-es.md)\r\n\r\n> **Secure execution layer for LLM/MCP agents** \u2014 safely integrate agents without exposing `.env`, secrets, or sensitive system files.\r\n\r\n---\r\n\r\n## \ud83d\udd0d What is Secure-MCP?\r\n\r\n**Secure-MCP** is a security-focused middleware framework that wraps around your MCP (Multi-Component Prompting) or LLM Agent systems to ensure that:\r\n\r\n- \ud83d\udeab `.env`, `API keys`, system files, and private configs are **not accessible** by agent logic.\r\n- \u2705 MCP/agent systems can still **safely perform operations**, such as calling APIs or retrieving tokens, via a secure proxy.\r\n- \ud83e\udde0 It supports LLM workflows, agent frameworks (like LangChain, AutoGen, CrewAI), and context-based tools \u2014 **without compromising security**.\r\n\r\n---\r\n\r\n## \ud83e\uddf1 Key Features\r\n\r\n| Feature                         | Description                                                                 |\r\n|--------------------------------|-----------------------------------------------------------------------------|\r\n| \ud83d\udd10 **SecureEnvBridge**          | Only whitelisted env variables can be accessed; `.env` is not directly exposed |\r\n| \ud83d\udcc2 **File Access Control**       | Blocks file reads like `.env`, `secret.pem`, `key.json`, `config.yaml` etc. |\r\n| \ud83d\udce6 **Secure Function Proxy**     | Sensitive operations are abstracted into secured callable functions         |\r\n| \ud83d\udcdc **Audit Logging**            | Logs all attempts to access sensitive resources                             |\r\n| \ud83e\uddea **Sandboxed Execution**       | Optionally runs MCP code in an isolated environment (e.g., Docker, WASM)     |\r\n| \u2699\ufe0f **Framework Agnostic**         | Easy to plug into any agent system or LLM pipeline                          |\r\n\r\n---\r\n\r\n## \ud83e\uddf0 Example Use Case\r\n\r\n```python\r\nfrom secure_mcp.secure_env_bridge import secure_get, secure_open\r\n\r\n# Only whitelisted environment variables are allowed\r\napi_key = secure_get(\"SERVICE_AUTH_TOKEN\")  # \u2705 allowed\r\ndb_pw = secure_get(\"DB_PASSWORD\")           # \u274c blocked\r\n\r\n# Safe file open (blocks sensitive files)\r\nwith secure_open(\"safe_data.csv\") as f:\r\n    content = f.read()\r\n```\r\n\r\nUse `SecureEnvBridge` instead of `os.getenv()` and `open()` to ensure safety.\r\n\r\n---\r\n\r\n## \ud83d\udd27 Installation\r\n\r\n```bash\r\ngit clone https://github.com/yourname/secure-mcp.git\r\ncd secure-mcp\r\npip install -e .\r\n```\r\n\r\nYou may also package it via `setup.py` or `pyproject.toml` for internal PyPI deployment.\r\n\r\n---\r\n\r\n## \ud83d\udee0\ufe0f Architecture\r\n\r\n```txt\r\n[ User Input / Agent Task ]\r\n           \u2193\r\n[ SecureContextExecutor ]\r\n           \u2193\r\n[ SecureEnvBridge / FileGuard ]\r\n           \u2193\r\n[ External API, File, or Memory Access ]\r\n           \u2193\r\n[ LLM Agent Response ]\r\n```\r\n\r\n* MCP doesn't need to know the secrets.\r\n* SecureEnvBridge acts like a **vault-aware delegate**.\r\n* Every access is **controlled, logged, and filtered**.\r\n\r\n---\r\n\r\n## \ud83d\udcc4 Example Structure\r\n\r\n```\r\nsecure-mcp/\r\n\u251c\u2500\u2500 secure_env_bridge/\r\n\u2502   \u251c\u2500\u2500 secure_get.py\r\n\u2502   \u251c\u2500\u2500 secure_open.py\r\n\u2502   \u251c\u2500\u2500 whitelist_loader.py\r\n\u2502   \u2514\u2500\u2500 audit_logger.py\r\n\u251c\u2500\u2500 examples/\r\n\u2502   \u251c\u2500\u2500 fastapi_demo.py\r\n\u2502   \u2514\u2500\u2500 sandbox_runner.py\r\n\u251c\u2500\u2500 tests/\r\n\u2502   \u2514\u2500\u2500 test_env_guard.py\r\n\u251c\u2500\u2500 .env.example\r\n\u251c\u2500\u2500 LICENSE\r\n\u251c\u2500\u2500 README.md\r\n\u2514\u2500\u2500 setup.py\r\n```\r\n\r\n---\r\n\r\n## \ud83d\udea7 Roadmap\r\n\r\n* [x] SecureEnvBridge with whitelist loader\r\n* [x] File access guard (`open()` patch)\r\n* [ ] WASM-based execution wrapper\r\n* [ ] Dockerized secure MCP runner\r\n* [ ] OpenAPI proxy adapter\r\n* [ ] Agent plugin system for LangChain / CrewAI\r\n\r\n---\r\n\r\n## \ud83d\udd10 License\r\n\r\nThis project is licensed under the **Apache License 2.0**.\r\nSee the [LICENSE](./LICENSE) file for more details.\r\n\r\n---\r\n\r\n## \ud83e\udd1d Contributing\r\n\r\nPull requests are welcome.\r\nFor major changes, please open an issue first to discuss what you would like to change.\r\nCheck `CONTRIBUTING.md` for detailed instructions.\r\n\r\n---\r\n\r\n## \ud83d\udcec Contact\r\n\r\nQuestions, ideas, or feedback?\r\nFeel free to [open an issue](https://github.com/yourname/secure-mcp/issues) or start a discussion.\r\n",
    "bugtrack_url": null,
    "license": "Apache-2.0",
    "summary": "Secure execution layer for LLM/MCP agents",
    "version": "0.1.0",
    "project_urls": {
        "Changelog": "https://github.com/yourname/secure-mcp/blob/main/CHANGELOG.md",
        "Documentation": "https://github.com/yourname/secure-mcp/wiki",
        "Homepage": "https://github.com/yourname/secure-mcp",
        "Issues": "https://github.com/yourname/secure-mcp/issues",
        "Repository": "https://github.com/yourname/secure-mcp"
    },
    "split_keywords": [
        "security",
        " mcp",
        " llm",
        " agent",
        " sandbox",
        " whitelist",
        " audit"
    ],
    "urls": [
        {
            "comment_text": null,
            "digests": {
                "blake2b_256": "28cf8562a0278173c910f2fd1fb1cb7b67930ccfbf127fb96d727cd2f11bf55f",
                "md5": "73e6cec2dbd9440be6aa8a4444148960",
                "sha256": "39f6a535f63ce6e6b0b7eba005d1bbcb5c5c8c267152ae557cb4c02ddaa8514a"
            },
            "downloads": -1,
            "filename": "secure_mcp-0.1.0-py3-none-any.whl",
            "has_sig": false,
            "md5_digest": "73e6cec2dbd9440be6aa8a4444148960",
            "packagetype": "bdist_wheel",
            "python_version": "py3",
            "requires_python": ">=3.7",
            "size": 10828,
            "upload_time": "2025-07-21T23:05:22",
            "upload_time_iso_8601": "2025-07-21T23:05:22.323799Z",
            "url": "https://files.pythonhosted.org/packages/28/cf/8562a0278173c910f2fd1fb1cb7b67930ccfbf127fb96d727cd2f11bf55f/secure_mcp-0.1.0-py3-none-any.whl",
            "yanked": false,
            "yanked_reason": null
        },
        {
            "comment_text": null,
            "digests": {
                "blake2b_256": "9e0124eaca6c50cb898993152350d643ff5164ecca8e9579971ea864cf8a84fe",
                "md5": "8bfc0c6404dbcc4bbacd82c76b217f9a",
                "sha256": "b931c102d557b333096cfd76ec7a7da9c4cc3cef9a29326cb99cc89e9719248d"
            },
            "downloads": -1,
            "filename": "secure_mcp-0.1.0.tar.gz",
            "has_sig": false,
            "md5_digest": "8bfc0c6404dbcc4bbacd82c76b217f9a",
            "packagetype": "sdist",
            "python_version": "source",
            "requires_python": ">=3.7",
            "size": 22957,
            "upload_time": "2025-07-21T23:05:23",
            "upload_time_iso_8601": "2025-07-21T23:05:23.661740Z",
            "url": "https://files.pythonhosted.org/packages/9e/01/24eaca6c50cb898993152350d643ff5164ecca8e9579971ea864cf8a84fe/secure_mcp-0.1.0.tar.gz",
            "yanked": false,
            "yanked_reason": null
        }
    ],
    "upload_time": "2025-07-21 23:05:23",
    "github": true,
    "gitlab": false,
    "bitbucket": false,
    "codeberg": false,
    "github_user": "re-rank",
    "github_project": "secure-mcp",
    "travis_ci": false,
    "coveralls": false,
    "github_actions": false,
    "requirements": [],
    "lcname": "secure-mcp"
}

re-rank