customgpt

This MCP server exposes a rich set of tools to interact with CustomGPT.ai agents, conversations, content, and API documentation from any MCP-compatible client (e.g., Claude Code, Claude Web). You can manage agents (list, get details, create from sitemaps or files, view statistics and settings), conduct conversations (send messages, list conversations, retrieve history), and manage agent content (list pages, check crawl/index status, manage sitemaps and uploads). Additional utilities include API key validation and an integrated API reference with endpoint details and examples. Use these tools by invoking the corresponding MCP commands with your API key and required identifiers (e.g., project_id) to perform tasks programmatically or via Claude-enabled interfaces.

Typical workflows include listing your agents, sending a message to an agent, and retrieving the chat history, or creating a new agent from a sitemap and then fetching its statistics. The server validates API keys, masks sensitive data in logs, and remains stateless with respect to API keys to enhance security during interactions with CustomGPT.ai.

Prerequisites:

• Python 3.10+ (as noted for FastMCP usage)
• Git
• An active CustomGPT.ai API key

Installation steps:

1. Clone the repository:

git clone https://github.com/Poll-The-People/customgpt-mcp.git
cd customgpt-mcp

2. (Optional) Create and activate a virtual environment:

python3.11 -m venv venv
source venv/bin/activate

3. Install dependencies:

pip install -r requirements.txt

4. Create or copy environment configuration and supply your API key:

cp .env.example .env
# Edit .env and set CUSTOMGPT_API_BASE, API_KEY, etc., as needed

5. Run the MCP server:

python server.py

Docker option:

git clone https://github.com/customgpt-ai/customgpt-mcp.git
cd customgpt-mcp
docker-compose up -d

Deployments (Railway, Vercel, etc.) follow the provider-specific setup in the repository README and require setting environment variables such as CUSTOMGPT_API_BASE, PORT, and PYTHONPATH.

Notes and tips:

• Ensure your CUSTOMGPT_API_BASE and API keys are set correctly in the environment (.env) or deployment settings.
• The MCP server runs on the configured HOST/PORT (default 0.0.0.0:8000 in the README example); adjust as needed for your environment and firewall rules.
• If using Docker, you can rely on docker-compose to manage dependencies and startup.
• The environment variable PYTHONPATH is often used to ensure the server imports resolve correctly when launched from MCP clients; set it to the project root if you structure imports that way.
• For Claude Web/Code integration, configure the MCP server URL and provide your CustomGPT API key within the tool configuration to enable authenticated requests.
• When handling API keys, the server masks keys in logs; avoid echoing keys in your scripts or logs.
• If you encounter port conflicts, change PORT in your environment and update any dependent client configurations accordingly.