Track every LLM user. Enforce limits directly in your code.
Tokenist is a Node.js module + REST API that gives you per-user metering, guardrails, and AI-based intent detection. Spot ToS breaches or jailbreak attempts, then block or rate-limit in the same workflow.
Free up to 10M tokens/month · No credit card required
Tokenist SDK flow
- 1.Import the @tokenist/guardrails Node module or call the REST endpoints from any backend.
- 2.Attach user identity (userId, orgId, feature) before you send a request to your LLM provider.
- 3.Call tokenist.check() to see if the user is allowed, then tokenist.record() after you get a response.
- 4.Use intent labels (frustration, jailbreak, ToS breach, etc.) to trigger automatic blocks or throttle logic.
Drop-in SDK + REST guardrails for your LLM stack
Tokenist plugs into your server-side code. You keep calling OpenAI (or any other provider) exactly as you do today — we just wrap each request with usage tracking, policy checks, and intent detection.
- Node-first, API everywhere — Use the TypeScript client when you're on Node, or hit the REST endpoints from any language/runtime.
- Per-user accounting — Usage, cost, and rules scoped to userId, orgId, or feature. Perfect for tiered plans and chargebacks.
- Intent-aware guardrails — Tokenist uses GPT-4o-mini to label every conversation — jailbreak attempt, ToS breach, frustration, win. Hook those signals into blocklists or custom automation.
- One rules engine — Set limits, send alerts, or block users from a single dashboard. No more scattered scripts or cron jobs.
The guardrails layer for every AI product
No proxies, no infra surgery. Just import the SDK (or call the REST API) and start tracking.
Node SDK + REST
Call `tokenist.check()` before your LLM request and `tokenist.record()` afterward. Works in Node via the TypeScript client or any stack via REST.
Per-user cost tracking
Get live token + cost data per user, org, and feature across every OpenAI model you use. Perfect for tiered plans, internal chargebacks, and budgeting.
Intent-aware guardrails
GPT-4o-mini labels every conversation (jailbreak, ToS breach, frustration, win). Use those signals to trigger automatic blocks, extra auth, or throttling.
Limits & automation
Create cost or token rules per user/tier, send Slack/webhook alerts, or auto-upgrade plans when someone hits a threshold.
Feature-level attribution
Tag each request with `feature` and see which product surfaces burn budget. Route expensive workflows to cheaper models before the bill arrives.
Block & throttle users
Add anyone to the blocklist (with optional expiry) or impose rolling 24h limits. Perfect for dealing with policy violators without killing the whole app.
Dashboards + API parity
Everything you can do in the dashboard (usage, limits, alerts) is available through the API. Automate whatever you don’t want to click through manually.
Audit-ready logs
Store request/response metadata for compliance and debugging. Opt-in to full payload logging when you need it, keep it lean when you don’t.
Use the SDK or hit the REST endpoints
Works with any backend stack. Wrap your OpenAI calls with Tokenist and you get metering, guardrails, and intent detection instantly.
import { tokenist } from '@tokenist/guardrails';
tokenist.init({ apiKey: process.env.TOKENIST_API_KEY });
// 1. Before you call OpenAI
await tokenist.check({
userId: 'user_abc123',
feature: 'voice-assistant',
model: 'gpt-4o-mini',
});
// 2. Call your provider as usual
const completion = await openai.responses.create({ ... });
// 3. Record usage + metadata (Tokenist handles cost calc)
await tokenist.record({
userId: 'user_abc123',
feature: 'voice-assistant',
usage: completion.usage,
sentiment: completion.response_text,
});curl https://api.tokenist.dev/sdk/check -H "Authorization: Bearer ug_..." -H "Content-Type: application/json" -d '{
"userId": "user_abc123",
"feature": "support-bot",
"model": "gpt-4o",
"requestType": "chat"
}'
# Response: { "allowed": true, "labels": ["safe"] }
# Use /sdk/record to log usage, /sdk/log for full payloads.Need Python, Go, or another runtime? Hit the REST endpoints directly — same rules engine, same intent labels, same dashboard visibility.
Simple pricing, based on requests
Pay for requests tracked. Start free, no credit card needed. We'll alert you before you hit your limit.
Free
Side projects and early evaluation
25,000 requests / mo
No credit card required
- Full enforcement — rate limits & blocklist
- AI intent labels (jailbreak, ToS breach)
- Basic dashboard + CSV export
- Community support
- No credit card required
Starter
Early commercial apps going to market
150,000 requests / mo
Then: $0.50 per 1,000 extra requests
- Everything in Free
- Email & Slack threshold alerts
- Per-feature cost attribution
- 30-day log retention
- Email support
Growth
Growing products with real traffic
750,000 requests / mo
Then: $0.30 per 1,000 extra requests
- Everything in Starter
- Webhook automations (e.g. auto-block on intent label)
- Cohort analysis + user segmentation
- 90-day log retention
- Priority support
Enterprise
1M+ requests / moCustom request volume, extended data retention, SSO + RBAC, dedicated SLA, and custom integrations — including on-prem options.
All paid plans include a 14-day free trial · Cancel any time · No credit card required to start
Common questions
Everything you need to know before getting started.