AIND Metadata access MCP server
LocalAbout AIND Metadata access MCP server
AIND Metadata access MCP server provides tools and resources for querying Allen Institute for Neural Dynamics metadata and data assets. It supports V1 metadata and is deprecated; the repository points users to a separate server for V2 metadata.
Key features
- 01Query MongoDB collections with filters
- 02Run MongoDB aggregation pipelines
- 03Explore schemas and documentation
- 04Load and inspect NWB files
- 05Generate data-asset summaries
Use cases
- 01Retrieve AIND metadata with filtered queries
- 02Explore metadata schemas and NWB files
- 03Summarize data assets for analysis
MCP support
Exposes AIND metadata and data-asset access through MCP tools and resources. It supports V1 metadata; the project is deprecated, and V2 metadata is handled by aind-data-mcp.
Configuration
Local server config is the same for every client. Copy it into your client’s MCP config file.
Tools
Local servers expose tools only once running on your machine, and the maintainer has not provided a tool list.
Details
From the project READMEArchive Notice
This repo is being archived.
AIND Metadata MCP Server
An MCP (Model Context Protocol) server that provides access to AIND (Allen Institute for Neural Dynamics) metadata and data assets through a comprehensive set of tools and resources.
Warning
This MCP server is deprecated. You can continue using it to access V1 metadata.
For V2 metadata, please see aind-data-mcp.
Features
This MCP server provides tools for:
- Data Retrieval: Query MongoDB collections with filters and projections
- Aggregation: Execute complex MongoDB aggregation pipelines
- Schema Exploration: Access detailed schema examples and documentation
- NWB File Access: Load and explore NWB (Neurodata Without Borders) files
- Data Summaries: Generate AI-powered summaries of data assets
Installation
Install uv if you haven't already - following this documentation
Install the MCP server using uv:
uv tool install aind-metadata-mcp
Or using pip:
pip install aind-metadata-mcp
Configuration
For Cline (VSCode Extension)
In order to ensure that the MCP server runs in your preferred client, you will have to download the aind-metadata-mcp package to your console. If space is an issue, please set UV_CACHE_DIR and UV_TOOL_DIR to locations that have capacity before proceeding with the next step.
-
Simpler version of install Run
uv tool install aind-metadata-mcpon your terminal and proceed below to configuring your MCP clients. -
If the above step didn't work:
Create virtual environment with python 3.11 in IDE
# Instructions for Conda
conda create -n <my_env> python=3.11
conda activate <my_env>
# Instructions for virtual environment
py -3.11 -m venv .venv
# Windows startup
.venv\Scripts\Activate.ps1
# Mac/ Linux startup
source .venv/bin/activate
Run the following commands in your IDE terminal.
pip install uv
uvx aind-metadata-mcp
If all goes well, and you see the following notice - Starting MCP server 'aind_data_access' with transport 'stdio'-, you should be good for the set up in your client of choice!
Cursor IDE
-
Install the MCP server:
uv tool install aind-metadata-mcp -
Create the MCP configuration file:
mkdir -p ~/.cursor -
Create
~/.cursor/mcp.jsonwith the following content:{ "mcpServers": { "aind-data-access": { "command": "aind-metadata-mcp", "args": [], "env": {} } } } -
Important: Replace
aind-metadata-mcpwith the full path if needed:which aind-metadata-mcpUse the full path (e.g.,
/Users/username/.local/bin/aind-metadata-mcp) in thecommandfield. -
Restart Cursor completely (Cmd+Q then reopen)
Note: Cursor uses a different MCP configuration system than VSCode. The configuration must be in ~/.cursor/mcp.json, not in the main settings.json file.
Instructions for use in MCP clients
JSON Config files to add MCP servers in clients should be structured like this
{
"mcpServers": {
}
}
Insert the following lines into the mcpServers dictionary
"aind_data_access": {
"disabled": false,
"timeout": 300,
"type": "stdio",
"command": "aind-metadata-mcp"
}
Note that after configuring the JSON files, it will take a few minutes for the serve to populate in the client.
Claude Desktop App
- Click the three lines at the top left of the screen.
- File > Settings > Developer > Edit config
Cline in VSCode
- Ensure that Cline is downloaded to VScode
- Click the three stacked rectangles at the top right of the Cline window
- Installed > Configure MCP Servers
- Close and reopen VSCode
Github Copilot in VSCode
- Command palette (ctr shift p)
- Search for MCP: Add server
- Select
Manual Install/stdio - When prompted for a command, input
uvx aind-data-access - Name your server
- Close and reopen VSCode
- In Copilot chat -> Select agent mode -> Click the three stacked rectangles to configure tools
- In order to enable the agent to reply with context of the AIND API, you'll have to manually add the .txt files (under resources) in this repository
For use in Code Ocean
- Locate the following capsule, to spin up Cline and Co-pilot with the aind-metadata-mcp pre-installed.
- Refer the the code ocean MCP server for additional support
- Either pin version 4.2, or 4.5
Comments
New to MCP?
From what the protocol is to installing your first server, these answers cover most first questions.
An open protocol that defines how AI applications connect to external tools and data in a uniform way. Clients talk to the model; MCP servers provide tools, resources and prompt templates. They exchange standard messages, so one server can be reused by many clients.