home / mcp / aind metadata mcp server

AIND Metadata MCP Server

Provides access to AIND metadata and data assets via MCP tools for retrieval, aggregation, schema exploration, NWB access, and data summaries.

Installation

Add the following to your MCP client configuration file.

Configuration

View docs

{
  "mcpServers": {
    "allenneuraldynamics-aind-metadata-mcp": {
      "command": "aind-metadata-mcp",
      "args": []
    }
  }
}

You can access AIND metadata and data assets through a dedicated MCP server that exposes data retrieval, aggregation, schema exploration, NWB file access, and AI-powered data summaries. This MCP server lets you connect with MCP clients to query, analyze, and explore data assets in a flexible, programmatic way.

How to use

Connect to the AIND Metadata MCP Server from your MCP client by configuring a local or remote server entry. Once connected, use the available capabilities to retrieve data, run aggregations, explore schemas, load NWB files, and generate summaries of assets.

How to install

Prerequisites you need to install and run the MCP server are Python 3.11 (for virtual environments), and the uv tool or pip for installation.

# Install the UV tool (preferred) and then install the MCP server
pip install uv
uv tool install aind-metadata-mcp

# Alternatively, install with pip directly
pip install aind-metadata-mcp

Additional configuration and notes

The MCP server runs as a local process and can be configured in MCP clients. A sample Cursor configuration shows how to start the server and point a client to it.

{
  "mcpServers": {
    "aind-data-access": {
      "command": "aind-metadata-mcp",
      "args": [],
      "env": {}
    }
  }
}

Cursor integration specifics

If you use Cursor, place the MCP configuration at ~/.cursor/mcp.json and restart Cursor after changes. The sample configuration above targets the aind-data-access MCP server.

Configuration details for client setup

Your MCP client will recognize the server as a stdio-based endpoint named aind_data_access. Use the exact command and arguments shown to start the server.

Available tools

Data Retrieval

Query MongoDB collections with filters and projections to retrieve specific data assets.

Aggregation

Execute complex MongoDB aggregation pipelines to derive insights and summarized results.

Schema Exploration

Access detailed schema examples and documentation to understand data structures.

NWB File Access

Load and explore NWB (Neurodata Without Borders) files for neuroscience data.

Data Summaries

Generate AI-powered summaries of data assets for quick understanding.