tsbridge

tsbridge acts as a tsnet-powered reverse proxy, letting you expose multiple backend services on your Tailnet from a single process. It's designed for homelabs and development environments where you want the magic of Tailscale without the hassle of running a separate sidecar for every service.

Inspired by Traefik, tsbridge can be configured with a simple TOML file or by watching Docker for container labels.

Why?

I got tired of spinning up a new tsnsrv instance for every service I wanted to expose on my Tailnet. Each one needs its own systemd service and configuration. With tsbridge, you configure once and add services as needed - either through a config file or by just adding labels to your Docker containers.

Install

Grab a binary from releases or:

go install github.com/jtdowney/tsbridge/cmd/tsbridge@latest

Quick start

1. Get OAuth credentials from https://login.tailscale.com/admin/settings/oauth
2. Create tsbridge.toml:

[tailscale]
oauth_client_id_env = "TS_OAUTH_CLIENT_ID"
oauth_client_secret_env = "TS_OAUTH_CLIENT_SECRET"

[[services]]
name = "api"
backend_addr = "127.0.0.1:8080"

[[services]]
name = "web"
backend_addr = "unix:///var/run/web.sock"

1. Run it:

export TS_OAUTH_CLIENT_ID=your-id
export TS_OAUTH_CLIENT_SECRET=your-secret
tsbridge -config tsbridge.toml

tsbridge will now be available on your tailnet. Thanks to MagicDNS, you can reach your services at https://api.<tailnet>.ts.net and https://web.<tailnet>.ts.net.

Configuration

tsbridge is configured via tsbridge.toml. See docs/quickstart.md for getting started quickly, or docs/configuration-reference.md for all options.

Here are a few common settings:

• whois_enabled: Set to true to add Tailscale-User-* identity headers to upstream requests
• write_timeout: Defaults to 30s. Set to "0s" to support long-running connections like Server-Sent Events (SSE)
• metrics_addr: Expose a Prometheus metrics endpoint (e.g., ":9090") - see docs/metrics.md for available metrics (secure this endpoint in production)

Security

Security Note: tsbridge is intended for homelabs and development environments. It hasn't been hardened or battle-tested for production workloads. See THREAT_MODEL.md for details.

For enhanced security using tag ownership, see Tag Ownership and OAuth Security.

Docker

Static config

docker run -v /path/to/config:/config \
  -e TS_OAUTH_CLIENT_ID=... \
  -e TS_OAUTH_CLIENT_SECRET=... \
  ghcr.io/jtdowney/tsbridge:latest -config /config/tsbridge.toml

Dynamic with labels (like Traefik)

tsbridge can watch Docker and automatically expose containers based on their labels:

# docker-compose.yml
services:
  tsbridge:
    image: ghcr.io/jtdowney/tsbridge:latest
    command: ["--provider", "docker"]
    volumes:
      - /var/run/docker.sock:/var/run/docker.sock # Required for label discovery
      - tsbridge-state:/var/lib/tsbridge
    environment:
      - TS_OAUTH_CLIENT_ID=${TS_OAUTH_CLIENT_ID}
      - TS_OAUTH_CLIENT_SECRET=${TS_OAUTH_CLIENT_SECRET}
    labels:
      - "tsbridge.tailscale.oauth_client_id_env=TS_OAUTH_CLIENT_ID"
      - "tsbridge.tailscale.oauth_client_secret_env=TS_OAUTH_CLIENT_SECRET"
      - "tsbridge.tailscale.state_dir=/var/lib/tsbridge"
      - "tsbridge.tailscale.default_tags=tag:server" # Must match or be owned by your OAuth client's tag

  whoami:
    image: traefik/whoami
    labels:
      - "tsbridge.enabled=true"
      - "tsbridge.service.name=whoami"
      - "tsbridge.service.port=80"

volumes:
  tsbridge-state:

See docs/docker-labels.md for the full label reference.

Note: The default_tags must match or be owned by your OAuth client's tag. Individual services can override this with their own tags label. See Tag Ownership and OAuth Security for setup details.

Headscale

Works with Headscale but requires auth keys instead of OAuth and TLS must be disabled (until headscale issue #2137 gets implemented):

[tailscale]
auth_key_env = "TS_AUTH_KEY"
control_url = "https://headscale.example.com"

[[services]]
name = "api"
backend_addr = "127.0.0.1:8080"
tls_mode = "off"  # Required for Headscale

See example/headscale/ for a complete setup.

Development

make build  # Build binary
make test   # Run tests
make lint   # Run linters

Monitoring

When metrics_addr is configured, tsbridge exposes Prometheus metrics at http://<metrics_addr>:

• Request counts and latencies
• Error rates
• Active connections
• Service lifecycle events

Architecture

Each service runs its own tsnet instance with isolated state, enabling independent lifecycle management and per-service configuration.

Prior Art

• tsnsrv - Inspiration but single service focused
• tsdproxy - Docker-specific

License

MIT License - see LICENSE file for details.