Not just what your
AI costs.
What it's worth.
The first platform that measures what your LLM calls produce, not just what they cost. Real-time margin visibility, pricing confidence, and six metrics the industry doesn't have yet.
No credit card · Free tier available at launch · 2 min setup
The industry has been
measuring the wrong thing.
Every LLM tool tracks input metrics. Tokens consumed. Cost per call. Requests per day. These tell you what you spent.
None of them tell you what you got. Tokflo introduces Token Intelligence — the discipline of measuring the return your AI investment produces.
Metrics that don't exist
anywhere else.
Tokflo defines, names, and publishes all six. First to name = first to own. Every developer discussion referencing these cites Tokflo.
Two agents consuming identical tokens — one drives $10K in outcomes, the other $200. VPT is 50× different. Every optimisation should be driven by VPT, not cost alone.
Not cost per call — cost per successful outcome. The "expensive" model that succeeds once often beats the "cheap" model that fails twice. TCC makes that visible.
Fires when a customer's usage trajectory will make them unprofitable by month end. You get days to act, not weeks. No other tool does this.
A single 0–100 score per agent. Above 70 = healthy. Below 40 = needs attention. Built for founders and investors, not just engineers.
When you change a prompt, did it improve return? PRI > 1.0 = v2 better. A prompt 30% costlier but 80% more effective has PRI 1.38 — worth it.
The minimum price for a plan to remain profitable. The average AI startup is undercharging by $30–80/mo per customer when they first see CPF.
5 minutes to integrate.
Instant intelligence.
No proxy. LLM calls go directly to providers. Tokflo wraps your client and ships events asynchronously — fire and forget.
import { Tokflo } from '@tokflo/sdk'
const tf = new Tokflo({ apiKey: process.env.TOKFLO_KEY })
// Wrap your existing client — 2 linesconst openai = tf.wrap(new OpenAI(), {
agent: 'invoice-parser',
customer: 'acme-corp',
prompt_version: 'v2.1'
})const result = await openai.chat.completions.create({...})
// Unlock VPT, TCC, PRI — 5 more linesawait tf.outcome({
status: 'completed',
value_usd: 50.00, user_accepted: true})Three intelligence layers
Connect Stripe once. Tokflo maps MRR per customer to enable margin and CPF calculation automatically.
One-time setupAdd 5 lines of SDK code per agent. Unlocks TCC, PRI, and precise VPT — the metrics that matter most.
5 lines of codeTokflo detects quality signals without tags: retry bursts, abandonment, output anomalies. Zero effort.
AutomaticEvery outcome tag trains our inference engine. After 10,000 tagged sessions, quality signals are detected without any code. A competitor starting 18 months later starts from zero.
Be first to know
what your AI is actually worth.
Tokflo is currently in private development. Join the waitlist to get priority access, early pricing, and direct input into what we build next.
Release roadmap
Closed alpha — SDK core + margin dashboard
Open beta — all 6 metrics + Stripe sync
Public launch — full Token Intelligence platform
Waitlist perks