{
  "id": 20734991,
  "title": "Private Tailnet Quotas for Small-Team LLM Gateways",
  "url": "https://urgent.news/2026/11/19/private-tailnet-quotas-for-small-team-llm-gateways",
  "topic": "ai",
  "section": "AI",
  "published": "2026-11-19T10:34:16.000Z",
  "source": {
    "name": "Dev.to",
    "slug": "dev-to",
    "url": "https://dev.to/james_lin/private-tailnet-quotas-for-small-team-llm-gateways-4hfb"
  },
  "original_language": "en",
  "account": "A small team using large language models (LLMs) requires a different approach to manage costs and access. Instead of a complex service mesh, they implement a simple solution with a single egress point for model calls. This egress is restricted to only the team's devices and has a hard limit on the number of tokens per user. Slack shared keys were previously used to enable this access, but are now replaced with a more secure method.\n\nTo implement this, a quota proxy runs on a small virtual machine (VM) within the team's private network, called a \"tailnet.\" The proxy intercepts requests from each device and checks if the user has enough tokens remaining for the request. If the limit is exceeded, the request is rejected before it leaves the network, preventing any unexpected and costly charges.\n\nThe proxy uses the Tailscale client, which is installed on each laptop and the gateway device. Each device gets a unique stable IP address and a MagicDNS name, allowing for easy access without opening any inbound ports. Access control is managed using Access Control Lists (ACLs) instead of traditional firewall rules.\n\nThe proxy communicates with the LLM provider via a reverse proxy, which verifies the user's identity and token quota before forwarding the request. The proxy is designed to be simple and reliable, with minimal dependencies and no logging of sensitive information. It uses a shared dictionary to keep track of token usage per user, but this can fail if multiple requests are made simultaneously. To mitigate this, an atomic increment and decrement mechanism could be implemented to ensure accurate tracking.\n\nThe solution is cost-effective, with a per-seat token cap of around $0.60 per day at a rate of $0.003 per 1,000 tokens. The team monitors their spending using a cron job that posts alerts to Slack when the estimated team spend crosses a certain threshold. The implementation balances security, simplicity, and cost control, making it an effective solution for small teams using LLMs.",
  "summary": "Private Tailnet Quotas for Small-Team LLM Gateways At 3:17 a.m. the page is the vendor dashboard. An agent loop on a shared key has spent the month, and the bill keeps climbing until someone revokes a credential copied into three laptops and a CI secret. A five-to-twenty person team does not need a service mesh. It needs one egress for model calls that only team devices can reach, and a hard…",
  "key_points": [
    "Small teams use LLMs, need cost management.",
    "Tailnet proxy restricts egress, token quotas per user.",
    "Slack keys replaced, Tailscale client enables secure access."
  ],
  "editors_take": "Replacing shared keys with a private proxy and token limits transforms LLM access from an open expense into a controlled, secure resource for small teams managing limited budgets.",
  "illustration": null,
  "coverage": {
    "outlets": 1,
    "also_reported_by": []
  },
  "ai_generated": true,
  "disclaimer": "Summaries, key points and the editor’s take are written by software from other outlets’ reporting and may contain errors — always check the linked original."
}