A Model Context Protocol (MCP) server that provides AI agents with access to Collibra Data Governance Center capabilities including data asset discovery, business glossary queries, and detailed asset information retrieval.
This Go-based MCP server acts as a bridge between AI applications and Collibra, enabling intelligent data discovery and governance operations through the following tools:
asset_details_get- Retrieve detailed information about specific assets by UUIDasset_keyword_search- Wildcard keyword search for assetsasset_types_list- List available asset typesbusiness_glossary_discover- Ask questions about terms and definitionsdata_classification_match_add- Associate a data class with an assetdata_classification_match_remove- Remove a classification matchdata_classification_match_search- Find associations between data classes and assetsdata_assets_discover- Query available data assets using natural languagedata_class_search- Search for data classes with filtersdata_contract_list- List data contracts with paginationdata_contract_manifest_pull- Download manifest for a data contractdata_contract_manifest_push- Upload manifest for a data contract
- Access to a Collibra Data Governance Center instance
- Valid Collibra credentials
-
Download the latest release:
- Go to the GitHub Releases page
- Download the appropriate binary for your platform:
chip-linux-amd64- Linux (Intel/AMD 64-bit)chip-linux-arm64- Linux (ARM 64-bit)chip-mac-amd64- macOS (Intel)chip-mac-arm64- macOS (Apple Silicon)chip-windows-amd64.exe- Windows (Intel/AMD 64-bit)chip-windows-arm64.exe- Windows (ARM 64-bit)
-
Make the binary executable (Linux/macOS):
chmod +x chip-* -
Optional: Move to your PATH:
# Linux/macOS sudo mv chip-* /usr/local/bin/mcp-server # Or add to your user bin directory mv chip-* ~/.local/bin/mcp-server
git clone <repository-url>
cd chip
go mod download
go build -o .build/chip ./cmd/chip
# Run the build binary
./.build/chipThe server supports two authentication approaches, either configured through environment variables or a configuration file
When running over the stdio transport, configure credentials at the server level - all requests use the same credentials:
mkdir -p ~/.config/collibra/Powershell:
New-Item -ItemType File -Path $HOME\.config\collibra\mcp.yamlbash/zsh:
touch ~/.config/collibra/mcp.yaml# ~/.config/collibra/mcp.yaml
api:
url: "https://your-collibra-instance.com"
username: "your-username"
password: "your-password"The same options can be configured through the respective environment variables COLLIBRA_MCP_API_URL, COLLIBRA_MCP_API_USR and COLLIBRA_MCP_API_PWD.
When running over the http transport, it is recommended that MCP clients provide their own Basic Auth headers for each request:
export COLLIBRA_MCP_API_URL="https://your-collibra-instance.com"
./mcp-serverFor detailed configuration instructions, see CONFIG.md.
- 🔐 Credentials: Store sensitive information in environment variables rather than config files
- 🌐 Network: HTTP mode binds to localhost only for security
- 🔒 TLS: Only use
skip-tls-verify: truefor development with self-signed certificates - 📁 File Permissions: Ensure config files have appropriate permissions when containing credentials
This server is compatible with any MCP client. Refer to your MCP client's documentation for server configuration.
Here's how to integrate with some popular clients assuming you have a configuration file setup:
- Claude desktop
// ~/Library/Application Support/Claude/claude_desktop_config.json
{
"mcpServers": {
"collibra": {
"type": "stdio",
"command": "/path/to/chip-..."
}
}
}- VS Code
// .vscode/mcp.json
{
"servers": {
"collibra": {
"type": "stdio",
"command": "./chip"
}
}
}- Gemini-cli
// ~/.gemini/settings.json
{
"mcpServers": {
"collibra": {
"command": "/path/to/chip-..."
}
}
}You can enable or disable specific tools by passing command line parameters, setting environment variables, or customizing the mcp.yaml configuration file.
You can specify tools to enable or disable by using the tool names listed above (e.g. asset_details_get). For more information, see the CONFIG.md documentation.
By default, all tools are enabled. Specifying tools to be enabled will enable only those tools. Disabling tools will disable only those tools and leave all others enabled. At present, enabling and disabling at the same time is not supported.