Documentation Index
Fetch the complete documentation index at: https://mintlify.com/InsForge/InsForge/llms.txt
Use this file to discover all available pages before exploring further.
Overview
The InsForge MCP (Model Context Protocol) server gives your AI coding assistant direct access to your backend infrastructure. Once connected, your AI can:- Fetch documentation - Get real-time docs for database, auth, storage, functions, and AI APIs
- Manage database schemas - Create tables, add columns, configure Row Level Security
- Query and inspect data - Get table schemas, view data, run custom SQL
- Manage storage - Create buckets, configure access policies
- Deploy functions - Create and update serverless edge functions
- Download templates - Scaffold new projects with InsForge pre-configured
Already using the VS Code extension? The InsForge VS Code extension provides one-click MCP installation.
Prerequisites
An AI coding assistant
InsForge MCP works with:
- Cursor - AI-first code editor
- Claude Code - Claude’s coding interface
- GitHub Copilot - In VS Code
- Windsurf, Cline, Roo Code, OpenCode - Alternative AI editors
- Google Antigravity, Codex, Kiro, Qoder - Emerging tools
An InsForge project
You need either:
- A cloud project at insforge.dev (easiest)
- A self-hosted instance running on Docker (see Deployment Guide)
Quick Setup (Recommended)
The fastest way to connect InsForge MCP:Option 1: One-Click Installation (Cloud)
If you’re using cloud-hosted InsForge:- Open your InsForge dashboard at insforge.dev
- Click Connect in the top navigation
- Select your AI assistant from the dropdown
- Click the Add to [Assistant] button or copy the installation command
- Follow the on-screen instructions
Option 2: Remote MCP (Universal)
Works with most AI assistants. One command handles everything:Supported clients: Claude Code, Claude Desktop, Codex, Cursor, Gemini CLI, Goose, GitHub Copilot, OpenCode, VS Code, Zed
- Automatically generates configuration for your AI client
- Prompts for authentication (opens browser)
- Binds to your InsForge project
Manual Configuration
Prefer to configure MCP manually? Choose your AI assistant:Cursor
Cursor
Installation
Method 1: One-Click (Cloud only)From your InsForge dashboard, click Add to Cursor to auto-generate the configuration.Method 2: Manual Configuration- Open Cursor Settings
- Navigate to Tools & MCP
- Click New MCP Server
- Paste the MCP JSON from your InsForge project settings
- Save the configuration
- Look for a green status indicator
MCP JSON Format
For local/self-hosted InsForge:Verify Installation
Start a new chat in Cursor and send:fetch-docs tool and return InsForge documentation.Claude Code
Claude Code
Installation
Method 1: Command-Line InstallationFrom your InsForge dashboard, select Claude Code and copy the installation command:.mcp.json in your home directory.Method 2: Manual ConfigurationCreate or edit ~/.mcp.json:Authentication
In your terminal (not the IDE), run:Verify Installation
Start a new chat in Claude Code and send:GitHub Copilot
GitHub Copilot
Windsurf
Windsurf
Cline
Cline
Installation (Smithery)
Install via Smithery CLI:Manual Installation
Follow the Cline documentation:- Open Cline Settings → MCP Servers
- Click Add Remote Server
- Enter:
- Server Name:
insforge - Server URL:
https://mcp.insforge.dev/mcp - Transport:
Streamable HTTP
- Server Name:
- Click Authenticate when prompted
Verify Installation
OpenCode
OpenCode
Codex
Codex
Other Clients (Generic MCP JSON)
Other Clients (Generic MCP JSON)
Universal MCP JSON
If your AI client supports MCP but isn’t listed above, use this generic configuration:For cloud-hosted InsForge (remote MCP):Getting Your API Key (Self-Hosted)
- Open your InsForge dashboard at
http://localhost:7130 - Navigate to Settings → API Keys
- Your API key starts with
ik_ - If not set, it’s auto-generated from your
.envfile underACCESS_API_KEY
Verify Installation
Available MCP Tools
Once connected, your AI assistant can use these tools:Documentation & Metadata
fetch-docs- Get SDK documentation for database, auth, storage, functions, AIget-backend-metadata- Retrieve your backend URL, anon key, API version
Database Management
run-raw-sql- Execute custom SQL queriesget-table-schema- Inspect table structure and columnslist-tables- Show all tables in your database
Storage Management
create-bucket- Create new storage bucketslist-buckets- Show all storage bucketsdelete-bucket- Remove storage buckets
Function Management
create-function- Deploy new serverless functionsupdate-function- Update existing functionsdelete-function- Remove deployed functionslist-functions- Show all deployed functions
Project Scaffolding
download-template- Download starter templates with InsForge pre-configured
See the MCP Tools Reference for detailed usage examples
Troubleshooting
Connection not working
Connection not working
Check MCP server status:In your AI assistant’s settings, verify the InsForge MCP server shows a green/connected status.For cloud-hosted: Ensure you completed the OAuth authentication flow.For self-hosted: Verify your API key is correct and starts with
ik_.Restart your AI assistant: Most clients require a full restart after MCP configuration changes.API key not found (self-hosted)
API key not found (self-hosted)
Check your If not set, InsForge auto-generates one on startup. Check the logs:Or open the dashboard at
.env file for ACCESS_API_KEY:http://localhost:7130 → Settings → API Keys.Changes not taking effect
Changes not taking effect
Most AI clients cache MCP configurations. To refresh:
- Close the AI assistant completely
- Restart the application
- Start a new chat session
Permission denied errors
Permission denied errors
Cloud-hosted: Re-authenticate using the OAuth flow. Your session may have expired.Self-hosted: Verify your API key has not been rotated. Check the dashboard for the current key.
Next Steps
Build Your First App
Create a todo app or Instagram clone in minutes
MCP Tools Reference
Complete list of MCP tools with examples
Core Features
Learn about database, auth, storage, and functions
Join Discord
Get help from the community