<div align="center">
<h1 align="center">
<br>
<a href="#"><img src="assets/COVER.png" alt="MCPStack Tool" width="100%"></a>
<br>
MCPStack MIMIC MCP
<br>
</h1>
<h4 align="center">Let Your Favourite LLM Dealing With The SQLs!</h4>
</div>
<div align="center">
<a href="https://pre-commit.com/">
<img alt="pre-commit" src="https://img.shields.io/badge/pre--commit-enabled-1f6feb?style=for-the-badge&logo=pre-commit">
</a>
<img alt="ruff" src="https://img.shields.io/badge/Ruff-lint%2Fformat-9C27B0?style=for-the-badge&logo=ruff&logoColor=white">
<img alt="python" src="https://img.shields.io/badge/Python-3.9%2B-3776AB?style=for-the-badge&logo=python&logoColor=white">
<img alt="pytest coverage" src="https://img.shields.io/badge/Coverage-65%25-brightgreen?style=for-the-badge&logo=pytest">
<img alt="license" src="https://img.shields.io/badge/License-MIT-success?style=for-the-badge">
</div>
> [!IMPORTANT]
> If you have not been across the MCPStack main orchestrator repository, please start
> there: [View MCPStack](https://github.com/MCP-Pipeline/MCPStack)
## <a id="about-the-project"></a>💡 About The MCPStack MIMIC Tool
`MCPStack MIMIC` is an MCP tool that connects the **MIMIC-IV clinical database** (with either SQLite or BigQuery backends)
into your **MCPStack pipelines**.
In layman's terms:
* MIMIC-IV is a large, de-identified database of ICU patient records, commonly used for healthcare research.
* This tool makes that dataset accessible to an LLM in a controlled way.
* It provides actions like *listing available tables*, *showing table structure with sample data*, and *running queries*; all exposed through MCP so your model can reason with healthcare data securely.
### What is MCPStack, in layman's terms?
The **Model Context Protocol (MCP)** standardises how tools talk to large language models.
`MCPStack` is the orchestrator: it lets you **stack multiple MCP tools together** into a pipeline and then expose them
inside an LLM environment (like Claude Desktop).
Think of it like **scikit-learn pipelines, but for LLMs**:
* In scikit-learn: you chain `preprocessors`, `transformers`, `estimators`.
* In MCPStack: you chain MCP tools (like MIMIC, Jupyter Notebook MCP, etc).
>[!IMPORTANT]
> This MCP has been made possible thanks to the `M3` original work by @rafiattrach, @rajna-fani, @MoreiraP12
> Under Dr. Leo Celi's supervision at MIT Lab for Computational Physiology. Following a first pull request of
> MCPStack to `M3`, we realised that we needed to externalise MCPStack to make it more modular and reusable
> across different use-cases. As such, MCPStack MIMIC is a copy of the original `M3` codebase, with adjustments
> only based on how how MCPStack works, and how it is structured.
---
## Installation
You can install the MIMIC tool as a standalone package. Thanks to `pyproject.toml` entry points, MCPStack
will auto-discover it.
### PyPI Installation Via `UV`
```bash
uv add mcpstack_mimic
```
### PyPI Installation Via `pip`
```bash
pip install mcpstack-mimic
```
### Install pre-commit hooks (optional, for development)
```bash
uv run pre-commit install
# or pip install pre-commit
```
### Run Unit Tests (optional, for development)
```bash
uv run pytest
```
---
## 🔌 Using With MCPStack
The `MIMIC` tool is auto-registered in MCPStack through its entry points:
```toml
[project.entry-points."mcpstack.tools"]
mimic = "mcpstack_mimic.tools.mimic.mimic:MIMIC"
```
That means MCPStack will “see” it without any extra configuration.
### Initialise the database
For SQLite (demo dataset by default):
```bash
uv run mcpstack tools mimic init --dataset mimic-iv-demo
```
This downloads and prepares the dataset locally.
### Configure the tool
Pick a backend (SQLite or BigQuery):
```bash
uv run mcpstack tools mimic configure --backend sqlite --db-path ./mimic.db
```
or
```bash
uv run mcpstack tools mimic configure --backend bigquery --project-id <YOUR_GCP_PROJECT>
```
This generates a `mimic_config.json` you can later feed into pipelines.
### Check status
```python
uv run mcpstack tools mimic status
```
>[!NOTE]
> We favourite `uv` for running MCPStack commands, but you can also use `mcpstack` directly if installed globally
> with `pip install mcpstack`.
---
## 🖇️ Build A Pipeline With MIMIC
Now that the tool is installed and configured, add it to your pipeline:
### Default MIMIC Pipeline (Runs with demo MIMIC dataset)
```bash
uv run mcpstack pipeline mimic --new-pipeline my_pipeline.json
```
### Create a new pipeline and add MIMIC previously custom-configured
```bash
uv run mcpstack pipeline mimic --new-pipeline my_pipeline.json --tool-config mimic_config.json
```
Or append to an existing pipeline
```bash
uv run mcpstack pipeline mimic --to-pipeline existing_pipeline.json --tool-config mimic_config.json
```
### Run it inside Claude Desktop
```bash
uv run mcpstack build --pipeline my_pipeline.json
```
Your LLM can now use the MIMIC tool in conversation, with secure access to the clinical dataset.
Open Claude Desktop, and tada!
---
## 📖 Programmatic API
```python
from mcpstack_mimic.tools.mimic.mimic import MIMIC
from mcpstack_mimic.tools.mimic.backend.backends.sqlite import SQLiteBackend
from mcpstack.stack import MCPStackCore
pipeline = (
MCPStackCore() #define =config if needed
.with_tool(MIMIC(
backends=[
SQLiteBackend("<path_to_your_mimic.db>") # SQLite backend with local MIMIC-IV database
])
# Here you can add as many as new `.with_tool(.)` of interest to play with.
).build(
type="fastmcp",
save_path="my_mimic_pipeline.json",
).run()
)
```
>[!IMPORTANT]
> The current repository has (1) technical debts, as in it would benefit from a refactor to make it maybe less messy; for instance, organising the actions into specific files. (2) a lack of documentation, the readme could deserve more in depth exploration of all the possible configurations, explore the code if you are a developer ; it is a little codebase.
> Pull Requests are more than welcome to minimise the tech debts and improve the documentation.
---
## 📽️ Video Demo
<video src="https://github.com/user-attachments/assets/e13c4a56-0fe9-41be-b789-407355fe1f4a" width="320" height="240" controls></video>
🔐 License
MIT — see **[LICENSE](LICENSE)**.
Raw data
{
"_id": null,
"home_page": null,
"name": "mcpstack-mimic",
"maintainer": null,
"docs_url": null,
"requires_python": ">=3.10",
"maintainer_email": null,
"keywords": "mcp, mcpstack, mimic-iv, tool, healthcare, bigquery, sqlite",
"author": null,
"author_email": "Provost Simon <s.g.provost@kent.ac.uk>",
"download_url": "https://files.pythonhosted.org/packages/93/5c/9c1c28c90f807cf3cb7543a48c165208fa1f06fb49f57cba6303fd949132/mcpstack_mimic-0.0.1.tar.gz",
"platform": null,
"description": "<div align=\"center\">\n <h1 align=\"center\">\n <br>\n <a href=\"#\"><img src=\"assets/COVER.png\" alt=\"MCPStack Tool\" width=\"100%\"></a>\n <br>\n MCPStack MIMIC MCP\n <br>\n </h1>\n <h4 align=\"center\">Let Your Favourite LLM Dealing With The SQLs!</h4>\n</div>\n\n<div align=\"center\">\n\n<a href=\"https://pre-commit.com/\">\n <img alt=\"pre-commit\" src=\"https://img.shields.io/badge/pre--commit-enabled-1f6feb?style=for-the-badge&logo=pre-commit\">\n</a>\n<img alt=\"ruff\" src=\"https://img.shields.io/badge/Ruff-lint%2Fformat-9C27B0?style=for-the-badge&logo=ruff&logoColor=white\">\n<img alt=\"python\" src=\"https://img.shields.io/badge/Python-3.9%2B-3776AB?style=for-the-badge&logo=python&logoColor=white\">\n<img alt=\"pytest coverage\" src=\"https://img.shields.io/badge/Coverage-65%25-brightgreen?style=for-the-badge&logo=pytest\">\n<img alt=\"license\" src=\"https://img.shields.io/badge/License-MIT-success?style=for-the-badge\">\n\n</div>\n\n> [!IMPORTANT]\n> If you have not been across the MCPStack main orchestrator repository, please start\n> there: [View MCPStack](https://github.com/MCP-Pipeline/MCPStack)\n\n## <a id=\"about-the-project\"></a>\ud83d\udca1 About The MCPStack MIMIC Tool\n\n`MCPStack MIMIC` is an MCP tool that connects the **MIMIC-IV clinical database** (with either SQLite or BigQuery backends)\ninto your **MCPStack pipelines**.\n\nIn layman's terms:\n* MIMIC-IV is a large, de-identified database of ICU patient records, commonly used for healthcare research.\n* This tool makes that dataset accessible to an LLM in a controlled way.\n* It provides actions like *listing available tables*, *showing table structure with sample data*, and *running queries*; all exposed through MCP so your model can reason with healthcare data securely.\n\n### What is MCPStack, in layman's terms?\n\nThe **Model Context Protocol (MCP)** standardises how tools talk to large language models.\n`MCPStack` is the orchestrator: it lets you **stack multiple MCP tools together** into a pipeline and then expose them\ninside an LLM environment (like Claude Desktop).\n\nThink of it like **scikit-learn pipelines, but for LLMs**:\n\n* In scikit-learn: you chain `preprocessors`, `transformers`, `estimators`.\n* In MCPStack: you chain MCP tools (like MIMIC, Jupyter Notebook MCP, etc).\n\n\n>[!IMPORTANT]\n> This MCP has been made possible thanks to the `M3` original work by @rafiattrach, @rajna-fani, @MoreiraP12\n> Under Dr. Leo Celi's supervision at MIT Lab for Computational Physiology. Following a first pull request of\n> MCPStack to `M3`, we realised that we needed to externalise MCPStack to make it more modular and reusable\n> across different use-cases. As such, MCPStack MIMIC is a copy of the original `M3` codebase, with adjustments\n> only based on how how MCPStack works, and how it is structured.\n\n---\n\n## Installation\n\nYou can install the MIMIC tool as a standalone package. Thanks to `pyproject.toml` entry points, MCPStack\nwill auto-discover it.\n\n### PyPI Installation Via `UV`\n\n```bash\nuv add mcpstack_mimic\n```\n\n### PyPI Installation Via `pip`\n\n```bash\npip install mcpstack-mimic\n```\n\n### Install pre-commit hooks (optional, for development)\n\n```bash\nuv run pre-commit install\n# or pip install pre-commit\n```\n\n### Run Unit Tests (optional, for development)\n\n```bash\nuv run pytest\n```\n\n---\n\n## \ud83d\udd0c Using With MCPStack\n\nThe `MIMIC` tool is auto-registered in MCPStack through its entry points:\n\n```toml\n[project.entry-points.\"mcpstack.tools\"]\nmimic = \"mcpstack_mimic.tools.mimic.mimic:MIMIC\"\n```\n\nThat means MCPStack will \u201csee\u201d it without any extra configuration.\n\n### Initialise the database\n\nFor SQLite (demo dataset by default):\n\n```bash\nuv run mcpstack tools mimic init --dataset mimic-iv-demo\n```\n\nThis downloads and prepares the dataset locally.\n\n### Configure the tool\n\nPick a backend (SQLite or BigQuery):\n\n```bash\nuv run mcpstack tools mimic configure --backend sqlite --db-path ./mimic.db\n```\n\nor\n\n```bash\nuv run mcpstack tools mimic configure --backend bigquery --project-id <YOUR_GCP_PROJECT>\n```\n\nThis generates a `mimic_config.json` you can later feed into pipelines.\n\n### Check status\n\n```python\nuv run mcpstack tools mimic status\n```\n\n>[!NOTE]\n> We favourite `uv` for running MCPStack commands, but you can also use `mcpstack` directly if installed globally\n> with `pip install mcpstack`.\n\n---\n\n## \ud83d\udd87\ufe0f Build A Pipeline With MIMIC\n\nNow that the tool is installed and configured, add it to your pipeline:\n\n### Default MIMIC Pipeline (Runs with demo MIMIC dataset)\n\n```bash\nuv run mcpstack pipeline mimic --new-pipeline my_pipeline.json\n```\n\n### Create a new pipeline and add MIMIC previously custom-configured\n\n```bash\nuv run mcpstack pipeline mimic --new-pipeline my_pipeline.json --tool-config mimic_config.json\n```\n\nOr append to an existing pipeline\n\n```bash\nuv run mcpstack pipeline mimic --to-pipeline existing_pipeline.json --tool-config mimic_config.json\n```\n\n### Run it inside Claude Desktop\n\n```bash\nuv run mcpstack build --pipeline my_pipeline.json\n```\n\nYour LLM can now use the MIMIC tool in conversation, with secure access to the clinical dataset.\nOpen Claude Desktop, and tada!\n\n---\n\n## \ud83d\udcd6 Programmatic API\n\n```python\nfrom mcpstack_mimic.tools.mimic.mimic import MIMIC\nfrom mcpstack_mimic.tools.mimic.backend.backends.sqlite import SQLiteBackend\nfrom mcpstack.stack import MCPStackCore\n\npipeline = (\n MCPStackCore() #define =config if needed\n .with_tool(MIMIC(\n backends=[\n SQLiteBackend(\"<path_to_your_mimic.db>\") # SQLite backend with local MIMIC-IV database\n ])\n # Here you can add as many as new `.with_tool(.)` of interest to play with.\n ).build(\n type=\"fastmcp\",\n save_path=\"my_mimic_pipeline.json\",\n ).run()\n)\n```\n\n>[!IMPORTANT]\n> The current repository has (1) technical debts, as in it would benefit from a refactor to make it maybe less messy; for instance, organising the actions into specific files. (2) a lack of documentation, the readme could deserve more in depth exploration of all the possible configurations, explore the code if you are a developer ; it is a little codebase.\n> Pull Requests are more than welcome to minimise the tech debts and improve the documentation.\n\n---\n\n## \ud83d\udcfd\ufe0f Video Demo\n\n<video src=\"https://github.com/user-attachments/assets/e13c4a56-0fe9-41be-b789-407355fe1f4a\" width=\"320\" height=\"240\" controls></video>\n\n\ud83d\udd10 License\n\nMIT \u2014 see **[LICENSE](LICENSE)**.\n",
"bugtrack_url": null,
"license": null,
"summary": "MIMIC-IV MCPStack tool (standalone package) with pluggable SQLite/BigQuery backends.",
"version": "0.0.1",
"project_urls": null,
"split_keywords": [
"mcp",
" mcpstack",
" mimic-iv",
" tool",
" healthcare",
" bigquery",
" sqlite"
],
"urls": [
{
"comment_text": null,
"digests": {
"blake2b_256": "bb4a99abb002a01a58f6864ee682bd34f46a652480224ca6a93b6e5437be6a8a",
"md5": "74fde777dffc6a27ae9ff809be8d9b65",
"sha256": "e1b08454cfc5db58c354eaaae638045d45d7bc504d53c7279ee196ce7b24a69c"
},
"downloads": -1,
"filename": "mcpstack_mimic-0.0.1-py3-none-any.whl",
"has_sig": false,
"md5_digest": "74fde777dffc6a27ae9ff809be8d9b65",
"packagetype": "bdist_wheel",
"python_version": "py3",
"requires_python": ">=3.10",
"size": 26853,
"upload_time": "2025-08-20T03:11:52",
"upload_time_iso_8601": "2025-08-20T03:11:52.906073Z",
"url": "https://files.pythonhosted.org/packages/bb/4a/99abb002a01a58f6864ee682bd34f46a652480224ca6a93b6e5437be6a8a/mcpstack_mimic-0.0.1-py3-none-any.whl",
"yanked": false,
"yanked_reason": null
},
{
"comment_text": null,
"digests": {
"blake2b_256": "935c9c1c28c90f807cf3cb7543a48c165208fa1f06fb49f57cba6303fd949132",
"md5": "28f99528f68b2ec0be65e01d7d270c84",
"sha256": "2839bd61a5615bf359ae0eeff5fbab8df061e6ec517520daff42e46077d86cb6"
},
"downloads": -1,
"filename": "mcpstack_mimic-0.0.1.tar.gz",
"has_sig": false,
"md5_digest": "28f99528f68b2ec0be65e01d7d270c84",
"packagetype": "sdist",
"python_version": "source",
"requires_python": ">=3.10",
"size": 29786,
"upload_time": "2025-08-20T03:11:54",
"upload_time_iso_8601": "2025-08-20T03:11:54.502250Z",
"url": "https://files.pythonhosted.org/packages/93/5c/9c1c28c90f807cf3cb7543a48c165208fa1f06fb49f57cba6303fd949132/mcpstack_mimic-0.0.1.tar.gz",
"yanked": false,
"yanked_reason": null
}
],
"upload_time": "2025-08-20 03:11:54",
"github": false,
"gitlab": false,
"bitbucket": false,
"codeberg": false,
"lcname": "mcpstack-mimic"
}
JSON