https://github.com/dnlbauer/cordra-mcp
A Model Context Protocol (MCP) server for enabling AI Assistants to interact with the Cordra Object Repository
https://github.com/dnlbauer/cordra-mcp
Last synced: 5 months ago
JSON representation
A Model Context Protocol (MCP) server for enabling AI Assistants to interact with the Cordra Object Repository
- Host: GitHub
- URL: https://github.com/dnlbauer/cordra-mcp
- Owner: dnlbauer
- License: mit
- Created: 2025-06-29T23:00:20.000Z (12 months ago)
- Default Branch: main
- Last Pushed: 2025-06-30T11:44:32.000Z (12 months ago)
- Last Synced: 2025-06-30T12:21:53.816Z (12 months ago)
- Language: Python
- Size: 108 KB
- Stars: 0
- Watchers: 0
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
# Cordra MCP Server
Cordra is a configurable digital object repository system that stores digital objects as JSON documents
with associated metadata and optional binary payloads.
This Model Context Protocol (MCP) server provides AI assistants with read-only
access to explore and understand Cordra repositories.
This allows AI systems to quickly understand the data model and schema structure
of a Cordra repository and to explore digital objects and their relationships.
## Features
- **Read-Only Access**: All operations are strictly read-only,
ensuring safe exploration without risk of data modification or corruption.
- **Schema Discovery**: Discover and retrieve schema definitions for each type in the repository.
- **Individual Object Retrieval**: Retrieve specific digital objects by their handle identifier with complete metadata.
## MCP Architecture
### Tools
- `list_types` - List all available types in the Cordra repository.
- Returns a JSON array of type names that are defined in the repository
- Types are returned in sorted order
- `get_type_schema` - Retrieve the JSON schema definition for a specific type.
- `type_name` - The name of the type (e.g., "Person", "Document", "Project")
- Returns the full schema definition as JSON
- `get_object` - Retrieve a digital object by its complete ID/handle.
- `object_id` - Complete object ID (e.g., "test/abc123")
- `search_objects` - Search for digital objects using a query string with pagination support.
- `query` - Lucene/Solr compatible search query
- `type` - Optional filter by object type
- `limit` - Number of results per page (default: 25)
- `page_num` - Page number to retrieve, 0-based (default: 0)
- `count_objects` - Count the total number of objects matching a query.
- `query` - Lucene/Solr compatible search query
- `type` - Optional filter by object type
- `get_design_object` - Retrieve the Cordra design object containing repository configuration.
- Includes type definitions, workflow configurations, and system settings
- Administrative privileges are typically required to access this object
#### Query Syntax
**CRITICAL**: JSON properties MUST be prefixed with `/`
✅ **Correct Examples:**
- `/title:*report*` - Wildcard search in title field
- `/author/name:Daniel` - Nested property access
- `/status:active AND /priority:high` - Boolean operators
- Use `type` parameter instead of including `type:` in query
❌ **Wrong (will fail):**
- `name:John` - Missing `/` prefix
- `author/name:Daniel` - Missing leading `/`
- `type:Person` - Use the `type` parameter instead
**Operators:** `*` (wildcard), `?` (single char), `AND`, `OR`, `NOT`, `"phrases"`
## Configuration
The MCP server can be configured using environment variables:
- `CORDRA_BASE_URL` - Cordra server URL (default: `https://localhost:8443`)
- `CORDRA_USERNAME` - Username for authentication (optional)
- `CORDRA_PASSWORD` - Password for authentication (optional)
- `CORDRA_VERIFY_SSL` - SSL certificate verification (default: `true`)
- `CORDRA_TIMEOUT` - Request timeout in seconds (default: `30`)
- `LOGLEVEL` - Logging level (default: `INFO`, options: `DEBUG`, `INFO`, `WARNING`, `ERROR`, `CRITICAL`)
## Usage
Run the MCP server:
```bash
uv run mcp-cordra
```
### Claude Code
You can add this MCP to Claude Code by registering it in the settings
of your project or creating a `.mcp.json` file in your workdir:
Example using the docker build:
```json
{
"mcpServers": {
"cordra": {
"command": "docker",
"args": [
"run",
"-i",
"--rm",
"-e",
"CORDRA_BASE_URL",
"ghcr.io/dnlbauer/cordra-mcp:latest"
],
"env": {
"CORDRA_BASE_URL": "https://cordra.example.de"
}
}
}
}
```
Example using the python package with `uvx`:
```json
{
"mcpServers": {
"cordra": {
"command": "uvx",
"args": [
"cordra-mcp"
],
"env": {
"CORDRA_BASE_URL": "https://cordra.example.de"
}
}
}
}
```