# Token Guardrail

> Token Guardrail is a developer tool that puts hard spend caps, live cost visibility, and spike alerts around the LLM API calls your application makes to Claude and GPT. It stops a runaway agent or a single abusive user from turning a $200 month into a $2,000 one — so you never find out about a blown budget from the invoice.

Token Guardrail is an SDK you wrap your existing Anthropic (Claude) and OpenAI (GPT) calls with, backed by a service that enforces budgets and a dashboard that shows where the money goes. It is aimed at developers and teams shipping AI features who need LLM cost control without rewriting their code.

## What it does

- **Hard spend caps**: Enforced budgets that stop calls at the limit — per project, per environment, and per month. Not advisory limits; when the cap is hit, calls are rejected.
- **Per-user quotas**: Budgets per individual user, so one person hammering your app cannot drain the whole budget.
- **Live cost visibility**: Spend as it happens, broken down by feature, model, and user — instead of waiting for the provider invoice.
- **Spike alerts**: Anomaly detection that pages you (e.g. in Slack) the moment spend departs from baseline — minutes in, not weeks later.
- **Drop-in SDK**: Wrap existing `anthropic.messages.create(...)` / OpenAI calls with the same interface; the calls are metered and capped with negligible latency.

## How it works

1. Wrap your calls — one line, the SDK drops in around your existing Claude or GPT calls.
2. Set the cap — per-project and per-user budgets; hard limits, not suggestions.
3. See it live — spend by feature and user, with alerts the moment something spikes.

## Who it is for

Developers and teams running LLM features in production who have been (or fear being) surprised by a sudden AI API bill: looping agents, retries that compound context, or a single user driving disproportionate cost.

## Status

Early access. Prospective users can join the waitlist on the homepage.

## Links

- Homepage: https://tokenguardrail.com
- GitHub: https://github.com/tokenguardrail
- X / Twitter: https://x.com/tokenguardrail