codehooks

This MCP server exposes Codehooks.io capabilities to AI agents that operate without a terminal. It provides access to database operations (collections, schemas, imports/exports), deployment of JavaScript serverless functions, file management (upload/list/delete metadata), a key-value store with TTL, and system operations like viewing logs and API docs. Use this server when the agent is running in Claude Desktop or similar environments where direct CLI access is not available. The MCP server sits behind Docker and uses the official RestDB Codehooks MCP image, so agents can trigger Codehooks actions via natural language prompts converted into MCP tool calls.

To use, configure the MCP in your desktop or terminal agent integration as described in the Setup section of the README. Once connected, the agent can perform tasks such as:

• Create and manage collections, schemas, and indexes; import/export data in JSON/JSONL/CSV formats.
• Deploy JavaScript serverless functions to Codehooks, including endpoints like POST /surveys or GET /surveys/:id/analytics.
• Upload and manage files in cloud storage, browse metadata, and delete files.
• Use a key-value store for settings and TTL-based data, with retrieval and deletion operations.
• View application logs and access local API documentation for the MCP agent.

Typical interactions will involve prompting the agent to create or update resources, deploy endpoints, or export data, after which the MCP will translate those prompts into Codehooks actions behind the scenes.

Prerequisites:

• Docker installed and running on the host.
• Access to the Codehooks Admin Token and project/spaces values (you will fill these in the env vars).
• Optional: Claude Desktop or another MC-powered client to load the MCP server in your configuration.

1. Create the MCP runner script (example shown for macOS/Linux). Save as codehooks.sh in a folder like ~/mcp-servers/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

2. Make the script executable:

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

3. If using Windows, create a batch file codehooks.bat with equivalent docker run command.

4. Configure the MCP in your client:

• Claude Desktop: add the config to claude_desktop_config.json with the path to the script, e.g.:

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

• Cursor: add the same entry to ~/.cursor/mcp.json, pointing to the script.

5. Environment values: replace the placeholders in the script or environment variables with your actual Codehooks project name, admin token, and space.

6. Start using the MCP: ensure Docker is running, then invoke the MCP via your client as documented in your integration guide.

Tips and common considerations:

• Replace the placeholder environment variables (CODEHOOKS_PROJECT_NAME, CODEHOOKS_ADMIN_TOKEN, CODEHOOKS_SPACE) with real values before running.
• If you rotate your admin token, update the environment variable and restart the MCP container.
• The MCP image is pulled automatically with --pull always; ensure you have network access to ghcr.io.
• If you see permission or path issues on macOS/Windows, ensure the script has executable permissions and that the path to the script is correct in your client config.
• This server is intended for environments without terminal access; for CLI-heavy workflows, using the Codehooks CLI directly is recommended (coho).