Connect your AI assistant to AIStackHub

Quick Start

Get your AI assistant connected in 2 minutes. Anonymous read-only access requires no token. Full access takes one extra step.

1. 1

   Add the MCP server to your client

   Server URL: https://aistackhub.ai/mcp — Add it to your client config. See the install guides below for client-specific instructions.

2. 2

   Test anonymous access (optional)

   Ask your AI: "Search the AIStackHub marketplace for the best AI writing tools under $50/mo" — no token needed for read-only tools.

3. 3

   Generate a token for full access

   Get a free token below or at the token generator to unlock stack building, assessments, and saved results.

4. 4

   Verify the connection

   Ask your AI: "Use AIStackHub to initialize a session and tell me what tools are available" — a successful response confirms the server is reachable.

Claude Desktop

Add AIStackHub to your Claude Desktop MCP configuration. Claude will automatically discover and use all 22 tools.

1. Locate your config file

Open the Claude Desktop config file:

open ~/Library/Application\ Support/Claude/claude_desktop_config.json

%APPDATA%\Claude\claude_desktop_config.json

~/.config/Claude/claude_desktop_config.json

2. Add the MCP server

{
  "mcpServers": {
    "aistackhub": {
      "type": "http",
      "url": "https://aistackhub.ai/mcp",
      "headers": {
        "Authorization": "Bearer YOUR_TOKEN_HERE"
      }
    }
  }
}

3. Restart Claude Desktop

Quit and relaunch Claude Desktop. You should see AIStackHub in the MCP tools list. Test with: "Search AIStackHub for AI coding tools"

Cursor

Add AIStackHub as an MCP server in Cursor via Settings.

1. 1

   Open Cursor Settings

   Press Cmd/Ctrl + Shift + J → navigate to MCP Servers tab.

2. 2

   Add server configuration

   Click "Add Server" and paste the config below.

3. 3

   Save and reload

   Cursor will detect all AIStackHub tools. Ask the Cursor AI agent to search the marketplace or build your stack.

{
  "mcpServers": {
    "aistackhub": {
      "url": "https://aistackhub.ai/mcp",
      "transport": "http-sse",
      "headers": {
        "Authorization": "Bearer YOUR_TOKEN_HERE"
      }
    }
  }
}

ChatGPT

Access AIStackHub capabilities in ChatGPT via a custom GPT Action. ChatGPT does not natively support the MCP protocol — use the REST API endpoints instead.

OpenAPI action snippet

{
  "openapi": "3.1.0",
  "info": { "title": "AIStackHub", "version": "1.0" },
  "servers": [{ "url": "https://aistackhub.ai/api/mcp" }],
  "paths": {
    "/marketplace/search": {
      "get": {
        "operationId": "marketplaceSearch",
        "summary": "Search AI tools marketplace",
        "parameters": [
          { "name": "q", "in": "query", "schema": { "type": "string" } },
          { "name": "limit", "in": "query", "schema": { "type": "integer" } }
        ]
      }
    }
  }
}

Any MCP-Compatible Client

AIStackHub uses MCP spec 2025-03-26 with Streamable HTTP/SSE transport. Any client that implements this spec can connect.

Get an API Token

A free API token unlocks write tools: stack building, readiness assessments, AI stack pulse, and saved results. Tokens are tied to your email, valid for 30 days, and renewable.

Tool Catalog

All 22 tools available on the AIStackHub MCP server. Tools marked anonymous require no token. Tools marked auth required need a valid API token.

Use Case Examples

Real conversations your AI agent can have using AIStackHub tools.

Auth & Scopes

AIStackHub uses scoped API tokens (ask_...) tied to verified email sessions. Tokens carry the same permissions as the browser session that generated them.

Token format

ask_7fK2mNpQrXvYzAbCdEfGhIjK  # ask_ prefix + 24 base62 chars

# Usage
Authorization: Bearer ask_7fK2mNpQrXvYzAbCdEfGhIjK

Scopes

Token lifecycle

Lifetime: 30 days. Tokens auto-renew on use (activity extends expiry). Inactive tokens expire hard at 30 days.

Revocation: Email hello@aistackhub.ai to revoke a token immediately. New token generation automatically revokes prior tokens for the same email.

Rate Limits

Limits are per token (or per IP for anonymous). Headers on every response show your current window.

X-RateLimit-Limit: 200
X-RateLimit-Remaining: 187
X-RateLimit-Reset: 1746662520  # Unix timestamp when window resets
Retry-After: 12                 # Only present on 429 responses

When rate-limited, the server returns HTTP 429. Implement exponential backoff starting at the Retry-After value.

Error Reference

All errors follow JSON-RPC 2.0 error format. The data.code field is the machine-readable error code.

{
  "jsonrpc": "2.0",
  "error": {
    "code": -32602,
    "message": "Invalid params",
    "data": {
      "code": "TOOL_NOT_FOUND",
      "message": "No tool with slug 'nonexistent-tool'"
    }
  },
  "id": 1
}