AI Agent Setup

A comprehensive guide for setting up AI coding agents to build Codehooks.io backend APIs. Covers the Claude Code plugin, CLI prompt, MCP server, OpenClaw skill, and manual setup for all major LLM agents.

What is Codehooks.io?

Codehooks.io is the agent-native backend platform.

It's a complete, isolated backend — API routes, NoSQL database, key-value store, worker queues, cron jobs, and static asset hosting — that deploys in under 5 seconds from a single CLI command. Everything is built in. Nothing needs to be wired together.

What makes it different is that the entire lifecycle is CLI-native:

• Create a project (coho create)
• Deploy code and assets (coho deploy)
• Inspect logs in real time (coho log -f)
• Query the database directly (coho query)
• Manage secrets, schemas, settings — all from the terminal

This means any tool that can run shell commands — Claude Code, Cursor, Codex, Cline, or a custom agent — can autonomously create, deploy, verify, and iterate on a production backend without human intervention.

For the developer, the job becomes: specify what you want, review what the agent built, and verify it works. Your agent handles the rest.

Getting Started

Choose the setup method that matches your AI coding tool:

• Claude Code Plugin — Recommended for Claude Code users. Auto-detection, templates, and full API context.
• CLI Prompt — For any agent with terminal access (Codex CLI, Cursor, Aider, Windsurf).
• MCP Server — Fallback for tools without terminal access (Claude Desktop, Cline, Zed). Prefer CLI when available.

Setup for AI Coding Agents

Claude Code Plugin (Recommended)

Best for: Claude Code users who want the fastest, most integrated experience

The Codehooks Claude Plugin gives Claude Code auto-detection of Codehooks projects, built-in templates, and a /codehooks:backend command for scaffolding production-ready backends from natural language.

What it provides:

• Auto-detection — Claude Code automatically recognizes Codehooks projects and loads platform context
• CLI integration — Deploy, query, and manage your backend directly from the conversation
• Templates — Pre-built patterns for webhooks, CRUD APIs, cron jobs, queues, and workflows
• /codehooks:backend command — Activate full backend development mode with a single slash command

Installation:

/plugin marketplace add RestDB/codehooks-claude-plugin
/plugin install codehooks@codehooks

Once installed, open any Codehooks project and use /codehooks:backend to start building.

Agents with Terminal Access

Best for: Codex CLI, Cursor, Aider, Windsurf, and other agents that can run shell commands

If your AI agent has terminal access, the Codehooks CLI provides the fastest setup with the prompt command:

# Install the CLI
npm install -g codehooks

# Display the prompt - pipe it to your agent or copy to clipboard
coho prompt

# Example: Copy to clipboard on macOS
coho prompt | pbcopy

The CLI also gives your agent direct access to deployment, database operations, logs, and more. See the CLI documentation for all available commands.

MCP-Compatible Tools

Best for: Claude Desktop, Cline, Zed, and other tools that support the Model Context Protocol but lack terminal access

For tools that support MCP, the Codehooks MCP server provides deploy, query, and management capabilities: https://github.com/RestDB/codehooks-mcp-server

Note: If your agent can run shell commands, always prefer the CLI over MCP — it's faster, more capable, and gives the agent direct access to the full coho toolchain.

Agent Skills & Templates

For AI agents that support skills or pre-built tool integrations, you can use the Codehooks OpenClaw Skill to get started even faster.

The codehooks-openclaw-skill provides pre-built patterns for common backend tasks:

• Webhook handlers with signature verification (Stripe, GitHub, etc.)
• CRUD REST APIs with validation
• Cron jobs for scheduled background work
• Worker queues for async processing
• Multi-step workflows with state management

Install the skill in your OpenClaw agent and it can scaffold production-ready Codehooks backends from natural language descriptions.

Read the full walkthrough in the blog post to see how OpenClaw agents use Codehooks as their serverless backend.

Development Prompt

The development prompt gives your AI agent full context on the Codehooks.io platform — API patterns, database operations, authentication, workers, cron, and more.

Access the prompt:

• CLI: coho prompt (pipe to your agent or copy to clipboard with coho prompt | pbcopy)
• Full docs: llms.txt — complete Codehooks documentation optimized for AI agents
• Source: codehooks-mcp-server — the canonical prompt is maintained here

For a walkthrough on using the prompt with ChatGPT, see the blog post.

Codehooks CLI for AI Agents

The Codehooks CLI (coho) is designed for AI agents working with command-line tools. Key commands for AI development:

Command	Description
coho prompt	Display the development prompt for AI assistants
coho docs	Open the API cheatsheet
coho deploy	Deploy code to the serverless cloud
coho log -f	Stream logs in real-time
coho query <collection>	Query database collections
coho info --examples	Show project info with cURL examples
coho doctor	Comprehensive project diagnostics as JSON — collections, stats, deploys, and recent error logs. Useful for agents to diagnose issues.
coho describe	Describe the deployed app structure — collections, schemas, queues, and files. Helps agents understand what's already deployed.

Full CLI documentation: https://codehooks.io/docs/cli

For additional detailed information about the Codehooks.io platform, you can reference https://codehooks.io/llms.txt