OpenAPI-tools

Within this repo I have OpenAPI-tools which you can use as MCP or directly with AI. It is inspired by open-webui/openapi-servers. I'm experimenting with different OpenAPI tools which I either consume directly with AI agents or convert into MCP servers.

Each server lives under servers/<name> and exposes a small FastAPI application that can run on its own or be incorporated into your automation pipelines.

Installation

To fetch the repository and set up a server in one step you can use the install scripts. They clone the repo from GitHub if needed or pull the latest changes before running the setup script.

Windows

PS> .\install.ps1 [path] [server_name]

Linux/WSL

$ ./install.sh [path] [server_name]

path is the directory where the repository will be cloned (defaults to OpenAPI-tools). server_name defaults to token_counter.

VS Code build task

The .vscode/tasks.json file contains a chain of tasks to help you run any server quickly:

1. Check for Virtual Environment – ensures a .venv directory exists.
2. Install Requirements – installs packages from the active server's requirements.txt.
3. Run API in localhost – starts the API using Uvicorn with the host and port defined in .vscode/settings.json.

Because these commands rely on PowerShell, the tasks currently work only on Windows. Linux and macOS users can run the same commands manually or adapt the tasks to their preferred shell.

To try it out, open a main.py file inside one of the server folders and press Ctrl+Shift+B. Once the build finishes, visit http://localhost:8282/docs (or your configured port) to explore the API.

Helper scripts

If you don't want to rely on the VS Code tasks, the repository also provides cross-platform setup and run scripts. Each script accepts an optional server name (defaults to token_counter).

Windows

PS> .\setup.ps1 [server_name]
PS> .\run.ps1 [server_name]

Linux/WSL

$ ./setup.sh [server_name]
$ ./run.sh [server_name]

setup creates the virtual environment, installs the requirements.txt packages and copies .env.example to .env if needed. run activates the environment and starts the API using the host and port from .vscode/settings.json (falling back to localhost:8000).

Servers

• token_counter – Simple API that counts tokens using tiktoken
• confluence_toolset_with_scope_restrictions – API for reading and writing Confluence Cloud pages

Design Notes

See docs/cors-design.md for background on why CORS is enabled on the example servers and how to tighten it for production deployments.