Stop overpaying for AI
you never meant to use.
Token Guardrail caps runaway LLM spend before it lands on your invoice - live tracking, hard limits, and instant alerts. Set a budget once; we hold the line.
7-day free trial · No credit card required
Trusted by teams shipping AI in production
- $4.2M+
- monthly spend under guardrail
- 18k+
- runaway spikes caught
- 99.98%
- cap-check uptime
- 5 min
- median time to first cap
Live cost visibility
See your spend the moment it happens
Not a monthly PDF. A live view of what your AI features actually cost - with the spike flagged and capped the instant it starts.
Spend today
6× baseline$284.10
Monthly cap
$284 / $500
Requests today
3,310
+12% vs. avg
Avg cost / req
$0.021
per request
Spend this month
Top spenders
by feature · today
Spike capped in #prod-app - chat-completion hit 6× baseline; calls paused at the $500 cap.
10:42 AMDrop-in SDK
Wrap the calls your app already makes
Same client, same interface - now metered and capped. No gateway to route through, no keys handed over. It's a wrapper, not a proxy.
import OpenAI from "openai";import { Guard } from "@tokenguardrail/sdk";// Wrap the client you already use - same interface, now metered.const openai = Guard(new OpenAI(), { cap: { monthly: 500 }, // hard limit, in USD alert: { slack: process.env.SLACK_WEBHOOK },});// Call it exactly like before. Token Guardrail meters every// request and stops the calls the moment the cap is hit.const res = await openai.chat.completions.create({ model: "gpt-4o-mini", messages: [{ role: "user", content: "Hello!" }],});Works with OpenAI, Anthropic, Google, Mistral, Groq - and any OpenAI-compatible SDK.
How it works
Three moves to take control of AI spend
Wrap the LLM calls your app already makes - same interface, now metered. From there you see it, cap it, and get alerted the moment something's off.
See every dollar
as it's spent
A live breakdown of AI spend by feature, model, and user. Spot the spike as it happens instead of finding out 30 days later.
- Per-request cost, attributed automatically
- Group by feature, model, or user
- Today vs. baseline, at a glance
How is this different from the usage limits my provider already offers?
Provider limits are account-wide and coarse. Token Guardrail caps by project, feature, and individual user, alerts you the moment spend departs from baseline, and shows exactly where the money goes - the control and visibility a provider dashboard doesn't give you.
One feature is driving most of the bill. Now you know which.
Budgets that stop the spend,
not just chart it
Set monthly caps per project and per user. As the budget runs low, fall back to cheaper models automatically; at the hard cap, calls stop with a clear error your app can handle.
- Enforced budgets, not advisory dashboards
- Automatic fallback to cheaper models
- A clean error to degrade gracefully
Can I set a spend cap per user?
Yes. You set budgets per project and per user, and the limits are enforced rather than advisory. As a user nears their budget you can automatically fall back to a cheaper model so they keep working, with the hard cap as the final backstop - so one person hammering your app can't drain the whole budget.
$284.10 used · 57% of cap
● enforced - calls stop at the cap
throw new CapExceededError()Get paged
before it hurts
Anomaly detection watches your baseline and pings Slack the moment spend departs from it - so you catch the loop while it's still cents, not thousands.
- Baseline anomaly detection
- Slack alerts in minutes
- Threshold and deviation triggers
How do I stop my OpenAI or Anthropic bill from spiking?
Token Guardrail wraps your existing Claude and GPT calls and enforces a hard spend cap. When a project or a single user crosses the budget, calls stop instead of quietly running up the bill - so a looping agent can't turn a $200 month into a $2,000 one.
Token GuardrailApp10:42 AM
Spend spike detected in #prod-app
⚠️ Runaway spend - action needed
Project
prod-app
Last 10 min
6× baseline
Driver
chat-completion
Security & privacy
Your keys and your content stay yours
Token Guardrail is a wrapper, not a gateway. It runs in your request path but keeps your credentials and payloads on your side.
Keys never leave your environment
The SDK wraps your provider client in your own backend. We never receive, store, or proxy your API keys.
We meter tokens, not your content
Token Guardrail measures cost - it doesn't warehouse your prompts or completions.
Local-first cap checks
Checks run against a cached budget, so your critical path never blocks on us.
Pricing
Pricing that scales with the spend you guard
Plans are priced on the LLM spend you put under guardrail - not on seats - so what you pay tracks the value it protects.
Starter
billed monthly
Up to $5k/mo monitored spend
For solo builders shipping their first AI feature.
- 3 projects
- Hard spend caps - project & per-user
- Spike alerts to Slack & email
- Graceful model fallback
- 14-day metric retention
Growth
billed monthly
Up to $25k/mo monitored spend
For teams that have been burned once already.
Everything in Starter, plus
- Unlimited projects
- Cost attribution by feature, model & user
- Trajectory insight - where the money leaked
- Per-user quotas
- 90-day retention
- Priority support
Scale
billed monthly
Up to $100k/mo monitored spend
For scaling teams that need control and guarantees.
Everything in Growth, plus
- SSO / SAML
- Audit logs
- Usage overages available
- 1-year retention
- SLA & priority support
Enterprise
Custom pricingFor organizations with security, compliance, and scale requirements. Unlimited monitored spend and a dedicated team.
- Unlimited monitored spend
- SSO / SCIM & security review
- DPA & custom retention
- Self-hosting / VPC option
- Dedicated support & onboarding
Every plan includes a free trial. No credit card required to start, and you can change or cancel anytime.
FAQ
Frequently asked questions
How do I stop my OpenAI or Anthropic bill from spiking?
Token Guardrail wraps your existing Claude and GPT calls and enforces a hard spend cap. When a project or a single user crosses the budget, calls stop instead of quietly running up the bill - so a looping agent can't turn a $200 month into a $2,000 one.
Can I set a spend cap per user?
Yes. You set budgets per project and per user, and the limits are enforced rather than advisory. As a user nears their budget you can automatically fall back to a cheaper model so they keep working, with the hard cap as the final backstop - so one person hammering your app can't drain the whole budget.
Does it work with both Claude and GPT?
Yes. The SDK drops in around your existing Anthropic and OpenAI calls with the same interface, so you keep your current models and code - the calls are simply metered and capped.
How is this different from the usage limits my provider already offers?
Provider limits are account-wide and coarse. Token Guardrail caps by project, feature, and individual user, alerts you the moment spend departs from baseline, and shows exactly where the money goes - the control and visibility a provider dashboard doesn't give you.
Will wrapping my calls add latency?
The guard check is a lightweight lookup that runs alongside your request, so the overhead is negligible next to the model's own response time.
What happens when a budget runs low?
You decide, in config. Declare a fallback ladder and, because the guard wraps your call, it swaps to the cheaper models you chose as the budget tightens - no routing code, and never a silent swap. Premium models restore when the budget resets. You can also have it queue requests, reject with a clear error, or just expose the budget state so you route it yourself. The hard cap is always the final backstop, so you never spend past the limit.
Do you store my prompts or completions?
Token Guardrail is built to meter tokens and cost, not to warehouse your content. Full data-handling details will ship with the privacy policy before general availability.
Which frameworks and SDKs does it support?
It wraps the official Anthropic and OpenAI SDKs directly, so it works whether you call them raw or through a framework like the Vercel AI SDK or LangChain. Support for more providers and integrations is on the roadmap.
How much will it cost?
Pricing isn't final yet. There will be a free tier for solo builders and paid plans for teams that need graceful model fallback, cost attribution, and more projects. Join the waitlist to lock in founding pricing.
When will Token Guardrail be available?
It's in early access. Join the waitlist and we'll reach out as spots open up.