Codehooks.io MCP Server

An MCP (Model Context Protocol) server that provides AI agents with database operations, serverless code deployment, and file management capabilities on the Codehooks.io platform.

When to Use This MCP vs. CLI Directly

Environment	Has Terminal?	Recommended Approach
Claude Code	Yes	Use `coho` CLI directly
Cursor	Yes	Use `coho` CLI directly
Claude Desktop	No	This MCP server

For AI Agents with Terminal Access (Claude Code, Cursor, etc.)

Using the Codehooks CLI directly is simpler and more flexible:

Install the CLI: npm install -g codehooks
Login: coho login
Get the AI prompt: coho prompt
Or reference the online docs: https://codehooks.io/llms.txt

The agent can then run coho commands directly in the terminal.

For Claude Desktop (No Terminal Access)

Use this MCP server to give the AI access to Codehooks functionality. See the Setup section below.

Available functionality

Database & Collections

Query and update collections (including metadata) with filters and sorting
Create and manage collections
Import/export data (JSON,JSONL,CSV)
Add schemas and indexes, cap collections

Code Deployment

Deploy JavaScript serverless functions

File Operations

Upload files to cloud storage
List and browse files
Delete files
Inspect file metadata

Key-Value Store

Store key-value pairs
Retrieve one or many key-value pairs
Delete key-value pairs
Set time-to-live (TTL) for key-value pairs

System Operations

View application logs
Access API documentation (local documentation for the MCP agent)

Setup

Get Codehooks Admin Token (keep it secret!)

coho login
coho add-admintoken

Create MCP Server Script

Create a folder for your MCP server scripts:

mkdir ~/mcp-servers
cd ~/mcp-servers

For macOS/Linux - Create codehooks.sh:

#!/bin/bash

# Set PATH to include common Docker locations
export PATH="/usr/local/bin:/opt/homebrew/bin:/usr/bin:/bin:$PATH"

exec docker run --rm -i \
  --pull always \
  -e CODEHOOKS_PROJECT_NAME=your_project_name \
  -e CODEHOOKS_ADMIN_TOKEN=your_admin_token \
  -e CODEHOOKS_SPACE=your_space_name \
  ghcr.io/restdb/codehooks-mcp:latest

Make it executable:

chmod +x ~/mcp-servers/codehooks.sh

For Windows - Create codehooks.bat:

@echo off
docker run --rm -i ^
  --pull always ^
  -e CODEHOOKS_PROJECT_NAME=your_project_name ^
  -e CODEHOOKS_ADMIN_TOKEN=your_admin_token ^
  -e CODEHOOKS_SPACE=your_space_name ^
  ghcr.io/restdb/codehooks-mcp:latest

Replace your_project_name, your_admin_token, and your_space_name with your actual values.

Configure for Claude Desktop

Add to your claude_desktop_config.json:

macOS/Linux:

{
  "mcpServers": {
    "codehooks": {
      "command": "/Users/username/mcp-servers/codehooks.sh"
    }
  }
}

Windows:

{
  "mcpServers": {
    "codehooks": {
      "command": "C:\\Users\\username\\mcp-servers\\codehooks.bat"
    }
  }
}

Configure for Cursor

Add to your ~/.cursor/mcp.json:

macOS/Linux:

{
  "mcpServers": {
    "codehooks": {
      "command": "/Users/username/mcp-servers/codehooks.sh"
    }
  }
}

Windows:

{
  "mcpServers": {
    "codehooks": {
      "command": "C:\\Users\\username\\mcp-servers\\codehooks.bat"
    }
  }
}

Replace username with your actual username.

Example Requests

"Build a complete survey system: create a database, deploy an API to collect responses, and add search/analytics endpoints"
"Set up a real-time inventory tracker: import my product CSV, create stock update webhooks, and build low-stock alerts"
"Build a webhook processing pipeline: receive webhooks from multiple sources, transform and validate data, then trigger automated actions"
"Build a content management system: create file upload endpoints, set up a metadata database, and deploy content delivery APIs"
"Set up automated data backups: export my collections to JSON files, store them with timestamps, and create restoration endpoints"

How These Examples Work

Complete Survey System

The AI agent would:

Create collections (surveys, responses) for data storage
Add schemas for data validation and structure
Deploy JavaScript endpoints like POST /surveys and GET /surveys/:id/analytics
Create indexes on response fields for fast searching and analytics

Real-time Inventory Tracker

The AI agent would:

Import your CSV to populate the products collection
Deploy webhook handlers for POST /inventory/update and GET /inventory/low-stock
Set up key-value storage for alert thresholds and settings
Create indexes on SKU and stock levels for real-time queries

Webhook Processing Pipeline

The AI agent would:

Deploy webhook receivers like POST /webhooks/stripe and POST /webhooks/github
Create collections for webhook_logs, processed_events, and failed_events
Set up data transformation rules and validation schemas for each webhook source
Use key-value store for rate limiting and duplicate detection with TTL
Deploy action triggers that send emails, update databases, or call other APIs based on webhook data

Content Management System

The AI agent would:

Create collections for content, media, and users
Deploy file upload endpoints with POST /upload and GET /content/:id
Upload and manage static files for content delivery
Store metadata linking files to content records with search indexes

Automated Data Backups

The AI agent would:

Export collections to JSON format with timestamps
Upload backup files to cloud storage automatically
Deploy restoration APIs like GET /backups and POST /restore/:backup-id
Store backup metadata in key-value store for tracking and management

Each example demonstrates how multiple MCP tools work together to create complete, production-ready systems through natural conversation with your AI agent.

Security Researchers

We thank the following individuals for responsible disclosure and helping improve the security of this project:

Liran Tal – Reported a command injection vulnerability in the query_collection tool (May 2025)

License

This project is licensed under the MIT License.

Codehooks.io MCP Server

An MCP (Model Context Protocol) server that provides AI agents with database operations, serverless code deployment, and file management capabilities on the Codehooks.io platform.

When to Use This MCP vs. CLI Directly

Environment	Has Terminal?	Recommended Approach
Claude Code	Yes	Use `coho` CLI directly
Cursor	Yes	Use `coho` CLI directly
Claude Desktop	No	This MCP server

For AI Agents with Terminal Access (Claude Code, Cursor, etc.)

Using the Codehooks CLI directly is simpler and more flexible:

Install the CLI: npm install -g codehooks
Login: coho login
Get the AI prompt: coho prompt
Or reference the online docs: https://codehooks.io/llms.txt

The agent can then run coho commands directly in the terminal.

For Claude Desktop (No Terminal Access)

Use this MCP server to give the AI access to Codehooks functionality. See the Setup section below.

Available functionality

Database & Collections

Query and update collections (including metadata) with filters and sorting
Create and manage collections
Import/export data (JSON,JSONL,CSV)
Add schemas and indexes, cap collections

Code Deployment

Deploy JavaScript serverless functions

File Operations

Upload files to cloud storage
List and browse files
Delete files
Inspect file metadata

Key-Value Store

Store key-value pairs
Retrieve one or many key-value pairs
Delete key-value pairs
Set time-to-live (TTL) for key-value pairs

System Operations

View application logs
Access API documentation (local documentation for the MCP agent)

Setup

Get Codehooks Admin Token (keep it secret!)

coho login
coho add-admintoken

Create MCP Server Script

Create a folder for your MCP server scripts:

mkdir ~/mcp-servers
cd ~/mcp-servers

For macOS/Linux - Create codehooks.sh:

#!/bin/bash

# Set PATH to include common Docker locations
export PATH="/usr/local/bin:/opt/homebrew/bin:/usr/bin:/bin:$PATH"

exec docker run --rm -i \
  --pull always \
  -e CODEHOOKS_PROJECT_NAME=your_project_name \
  -e CODEHOOKS_ADMIN_TOKEN=your_admin_token \
  -e CODEHOOKS_SPACE=your_space_name \
  ghcr.io/restdb/codehooks-mcp:latest

Make it executable:

chmod +x ~/mcp-servers/codehooks.sh

For Windows - Create codehooks.bat:

@echo off
docker run --rm -i ^
  --pull always ^
  -e CODEHOOKS_PROJECT_NAME=your_project_name ^
  -e CODEHOOKS_ADMIN_TOKEN=your_admin_token ^
  -e CODEHOOKS_SPACE=your_space_name ^
  ghcr.io/restdb/codehooks-mcp:latest

Replace your_project_name, your_admin_token, and your_space_name with your actual values.

Configure for Claude Desktop

Add to your claude_desktop_config.json:

macOS/Linux:

{
  "mcpServers": {
    "codehooks": {
      "command": "/Users/username/mcp-servers/codehooks.sh"
    }
  }
}

Windows:

{
  "mcpServers": {
    "codehooks": {
      "command": "C:\\Users\\username\\mcp-servers\\codehooks.bat"
    }
  }
}

Configure for Cursor

Add to your ~/.cursor/mcp.json:

macOS/Linux:

{
  "mcpServers": {
    "codehooks": {
      "command": "/Users/username/mcp-servers/codehooks.sh"
    }
  }
}

Windows:

{
  "mcpServers": {
    "codehooks": {
      "command": "C:\\Users\\username\\mcp-servers\\codehooks.bat"
    }
  }
}

Replace username with your actual username.

Example Requests

"Build a complete survey system: create a database, deploy an API to collect responses, and add search/analytics endpoints"
"Set up a real-time inventory tracker: import my product CSV, create stock update webhooks, and build low-stock alerts"
"Build a webhook processing pipeline: receive webhooks from multiple sources, transform and validate data, then trigger automated actions"
"Build a content management system: create file upload endpoints, set up a metadata database, and deploy content delivery APIs"
"Set up automated data backups: export my collections to JSON files, store them with timestamps, and create restoration endpoints"

How These Examples Work

Complete Survey System

The AI agent would:

Create collections (surveys, responses) for data storage
Add schemas for data validation and structure
Deploy JavaScript endpoints like POST /surveys and GET /surveys/:id/analytics
Create indexes on response fields for fast searching and analytics

Real-time Inventory Tracker

The AI agent would:

Import your CSV to populate the products collection
Deploy webhook handlers for POST /inventory/update and GET /inventory/low-stock
Set up key-value storage for alert thresholds and settings
Create indexes on SKU and stock levels for real-time queries

Webhook Processing Pipeline

The AI agent would:

Deploy webhook receivers like POST /webhooks/stripe and POST /webhooks/github
Create collections for webhook_logs, processed_events, and failed_events
Set up data transformation rules and validation schemas for each webhook source
Use key-value store for rate limiting and duplicate detection with TTL
Deploy action triggers that send emails, update databases, or call other APIs based on webhook data

Content Management System

The AI agent would:

Create collections for content, media, and users
Deploy file upload endpoints with POST /upload and GET /content/:id
Upload and manage static files for content delivery
Store metadata linking files to content records with search indexes

Automated Data Backups

The AI agent would:

Export collections to JSON format with timestamps
Upload backup files to cloud storage automatically
Deploy restoration APIs like GET /backups and POST /restore/:backup-id
Store backup metadata in key-value store for tracking and management

Each example demonstrates how multiple MCP tools work together to create complete, production-ready systems through natural conversation with your AI agent.

Security Researchers

We thank the following individuals for responsible disclosure and helping improve the security of this project:

Liran Tal – Reported a command injection vulnerability in the query_collection tool (May 2025)

License

This project is licensed under the MIT License.