Tried `tailscale/tailcat`: A Small Tool with Big Private-Network Potential

Tailscale's new tool, tailcat, offers netcat-like functionality over Tailscale's data plane without control-plane dependency, gaining over 790 GitHub stars in a day. Developers can use it to securely route TCP traffic to private AI gateways, avoiding public exposure and reducing VPN dependencies. The tool is positioned as a lightweight solution for teams building private AI infrastructure, with recommendations to keep quota enforcement and key rotation at the gateway level.

tailscale/tailcat : A Small Tool with Big Private-Network Potential tailscale/tailcat is essentially netcat over Tailscale’s data plane—without depending on Tailscale’s control plane. That makes it interesting for teams building private AI infrastructure: you can move TCP traffic between trusted nodes without exposing an inference gateway publicly or introducing another VPN control dependency. The project is currently seeing strong momentum, with +790 GitHub stars today . The appeal is easy to understand: it is small, composable, and fits the way infrastructure engineers already think about nc , port forwarding, and service probes. A practical architecture is to keep the AI gateway on a private Docker network, then expose only the required listener through Tailcat: docker-compose.yml services: ai-gateway: image: your-openai-compatible-gateway:latest environment: OPENAI BASE URL: https://b-lost.com/v1 OPENAI API KEY: ${BLOST API KEY} DEFAULT MODEL: claude-fable-5 REQUEST LOGGING: "false" ports: - "127.0.0.1:8080:8080" From a trusted build runner or developer workstation, the gateway can then be reached over the Tailcat transport according to the repository’s listener/connect syntax: curl http://private-gateway.internal:8080/v1/chat/completions \ -H "Authorization: Bearer $TEAM GATEWAY KEY" \ -H "Content-Type: application/json" \ -d '{ "model": "claude-fable-5", "messages": {"role":"user","content":"Review this deployment plan."} }' For enterprise use, I would place quota enforcement, key rotation, and model allowlists in the gateway—not in Tailcat. Tailcat should stay focused on private transport. Keep local access logs disabled or redacted, and avoid sending sensitive payloads through unnecessary observability layers. If using B-Lost as the upstream relay, its OpenAI-compatible base URL is https://b-lost.com/v1 . Native Anthropic /v1/messages prompt caching can also reduce repeated-context costs, with cache hits discounted by 90%. That combination—private routing plus caching—helps keep both latency and token spend predictable without weakening team-level governance.