caddytail 0.5.1

Caddy web server with Tailscale plugin, packaged for pip installation

caddytail

Caddy web server with the Tailscale plugin, packaged for pip installation. Run any Python web app on your tailnet with one command — Flask, FastAPI, Django, or any WSGI/ASGI callable.

Installation

pip install caddytail              # bare WSGI apps — no extra deps
pip install caddytail[flask]       # Flask support
pip install caddytail[fastapi]     # FastAPI / Starlette support

Quick Start

Write a normal Flask app — no CaddyTail-specific setup needed:

# app.py
from flask import Flask
from caddytail import get_user

app = Flask(__name__)

@app.get("/")
def index():
    user = get_user()
    return f"Hello, {user.name}!"

Or use any WSGI callable — no framework required:

# app.py
from caddytail import get_user

def app(environ, start_response):
    user = get_user(environ)
    body = f"Hello, {user.name}!" if user else "Not authenticated"
    start_response("200 OK", [("Content-Type", "text/plain")])
    return [body.encode()]

Run it on your tailnet:

caddytail run myapp app:app

That's it. Your app is now available at https://myapp.<tailnet>.ts.net with Tailscale authentication.

CLI

Hostname is always the first positional argument:

# Development — foreground, Ctrl-C kills everything
caddytail run <hostname> <app_ref> [--debug]

# Production — install as systemd service + tail logs
caddytail install <hostname> <app_ref> [--no-start] [--env K=V]

# Service management
caddytail status <hostname>
caddytail logs <hostname> [-n LINES] [-f]
caddytail restart <hostname>
caddytail uninstall <hostname>

# List all installed services
caddytail list

# Raw Caddy pass-through
caddytail caddy [args...]

The <app_ref> format is module:variable (like uvicorn), defaulting the variable to app:

• app:app — import app from app.py
• myproject.main:application — import application from myproject/main.py
• app — shorthand for app:app

Behavior

• run — starts Caddy + your app in the foreground. Ctrl-C kills everything. The framework is auto-detected: Flask and FastAPI get framework-specific middleware; generic WSGI apps are served with wsgiref; generic ASGI apps are served with uvicorn.
• install — writes a systemd unit file (ExecStart = caddytail run ...), enables, starts. If stdout is a tty, automatically tails logs. Ctrl-C stops tailing but leaves the service running.
• uninstall — stops, disables, and removes the unit file.
• caddy — passes all remaining args to the bundled Caddy binary.

Python API

get_user()

Returns a TailscaleUser with .name, .login, .profile_pic:

from caddytail import get_user

# Flask — no arguments needed (uses flask.request automatically)
user = get_user()

# FastAPI / Starlette — pass the Request object
user = get_user(request)

# WSGI — pass the environ dict
user = get_user(environ)

# Django — pass request.META
user = get_user(request.META)

if user:
    print(user.name)        # "John Doe"
    print(user.login)       # "[e-mail]"
    print(user.profile_pic) # "https://..."

login_required

Works as both a Flask decorator and a FastAPI Depends() target:

from caddytail import login_required

# Flask
@app.get("/secret")
@login_required
def secret():
    user = get_user()
    return f"Hello, {user.name}!"

# FastAPI
@app.get("/secret")
async def secret(user=Depends(login_required)):
    return {"message": f"Hello, {user.name}!"}

static()

Register static file paths to be served directly by Caddy:

from caddytail import static

static(app, "/assets/*", "./static")
static(app, "/uploads/*", "/var/www/uploads")

The runner picks these up automatically when starting Caddy.

CaddyTail class

For programmatic use (most users should use the CLI runner instead):

from caddytail import CaddyTail

caddy = CaddyTail(app, "myapp", debug=True)
caddy.run()

All ports are auto-allocated. No conflicts when running multiple apps.

Framework Examples

FastAPI

from fastapi import FastAPI, Request, Depends
from caddytail import get_user, login_required

app = FastAPI()

@app.get("/")
async def index(request: Request):
    user = get_user(request)
    return {"message": f"Hello, {user.name}!"}

@app.get("/protected")
async def protected(user=Depends(login_required)):
    return {"message": f"Hello, {user.name}!"}

Django

# views.py
from django.http import HttpResponse
from caddytail import get_user

def index(request):
    user = get_user(request.META)
    return HttpResponse(f"Hello, {user.name}!")

Bare WSGI

from caddytail import get_user

def app(environ, start_response):
    user = get_user(environ)
    body = f"Hello, {user.name}!" if user else "Not authenticated"
    start_response("200 OK", [("Content-Type", "text/plain")])
    return [body.encode()]

ASGI

from caddytail import get_user

async def app(scope, receive, send):
    # For ASGI apps, extract headers from the scope manually
    ...

All examples are run the same way:

caddytail run myapp myproject:app

Supported Platforms

Pre-built wheels are available for:

Platform	Architecture
Linux (glibc)	x86_64, aarch64
macOS	x86_64 (Intel), arm64 (Apple Silicon)
Windows	x86_64

Building from Source

git clone https://github.com/jpc/caddytail
cd caddytail

# Install Go and xcaddy
go install github.com/caddyserver/xcaddy/cmd/xcaddy@latest

# Build caddy with the tailscale plugin
xcaddy build --with github.com/tailscale/caddy-tailscale --output src/caddytail/bin/caddy

# Build the wheel
pip install build
python -m build --wheel

License

This project packages Caddy (Apache 2.0 License) with the Tailscale plugin (BSD 3-Clause License).

Links

• Caddy
• Tailscale
• caddy-tailscale plugin
• xcaddy