- Home
- MCP servers
- Directus
Directus
- javascript
8
GitHub Stars
javascript
Language
6 months ago
First Indexed
2 months ago
Catalog Refreshed
Documentation & install
Readme and setup notes from the catalogue, plus a client-ready config you can copy for your MCP host.
The Directus MCP Server Extension lets any MCP client connect to your Directus instance, enabling content editors to manage content and data analysts to query collections through natural language. It provides safe, streaming HTTP communication and a curated set of tools to interact with Directus without risking destructive actions.
How to use
You connect an MCP client to your Directus instance using either a remote URL configuration or a local, npx-based bridge. Once connected, you can browse your Directus schema, read collections and items, manage files and fields, run automation flows, and interact with content through natural language prompts. The server limits destructive actions such as deleting fields or collections to protect your data while still offering powerful reading, creating, updating, and commenting capabilities.
How to install
Prerequisites you need before starting:
- A Directus project ready to use
Install Directus extension and prepare credentials as described below.
Additional sections
MCP connection methods are provided for HTTP remote access and for local bridging via a command runner. The HTTP configuration uses a shared URL, while the local bridge uses a command that runs an MCP bridge client (for clients that lack native remote support). You can structure your setup to immediately connect and begin querying Directus data.
System prompts and prompts configuration
You can customize the system prompt for guiding the LLM’s tool usage by setting MCP_SYSTEM_PROMPT in your Directus environment. You can also disable it by setting MCP_SYSTEM_PROMPT_ENABLED to false. For prompts stored inside Directus, enable the prompts collection and set the corresponding field mappings to define names, descriptions, system prompts, and messages. You can use mustache templating to substitute variables in prompts with values you supply at call time.
Security and usage notes
The extension is designed to be safe by default, avoiding destructive changes. Always protect access tokens and use HTTPS URLs to prevent token leakage.
Tools exposed by the MCP server
The MCP server provides a comprehensive set of tools to interact with Directus data and metadata, including reading schema, reading and creating items, updating fields, managing files, and running automations. This enables you to perform content operations, data analysis, and workflow actions from MCP-enabled clients.
Available tools
system-prompt
Provides context to the LLM about its role and session start, guiding tool usage
users-me
Fetches current user information to tailor responses and permissions
read-collections
Retrieves the schema of all collections to explore structure and relationships
read-items
Fetches items from a collection for display or searching
create-item
Creates new items in collections to add content or records
update-item
Updates existing items to modify content or statuses
delete-item
Removes items from collections to clean up outdated content
read-files
Accesses file metadata or content to locate assets
import-file
Imports files from URLs to add external media
update-files
Updates file metadata for organization and tagging
read-fields
Gets field definitions for collections to understand data structure
read-field
Gets specific field information for detailed configuration
create-field
Adds new fields to collections to extend data models
update-field
Modifies existing fields including validation and UI options
read-flows
Lists available automation flows for discovery
trigger-flow
Executes automation flows for bulk operations or status changes
read-comments
Views comments on items to retrieve feedback and discussions
upsert-comment
Adds or updates comments to document decisions or feedback
markdown-tool
Converts between markdown and HTML for WYSIWYG fields
get-prompts
Lists available prompt templates for reuse
get-prompt
Executes a stored prompt template for consistent interactions