Route AI prompts to the best model, best price.
Intelligent middleware that analyzes intent, applies semantic caching, and routes to the optimal LLM. Cut costs by up to 90%.
1 · Prompt Composer
Normalizes the raw prompt — applies your presets, redacts PII, and checks the semantic cache before a single token is spent. Click any stage to pin its details.
Routing across 53 models from 18 brands
Everything you need
A complete middleware solution for intelligent AI routing.
Intelligent Routing
AI-powered intent analysis routes prompts to the optimal model based on complexity, speed, and cost requirements.
Cost Arbitrage
Save up to 90% on API costs by automatically selecting the most cost-effective provider for each request.
Semantic Caching
85% similarity matching eliminates redundant API calls. Zero-cost responses for semantically identical queries.
Security First
Built-in PII redaction, threat detection, and comprehensive audit logging for enterprise compliance.
Cascading Fallbacks
Automatic failover across 53 models from 18 brands ensures 99.95% uptime for your applications.
Latency Optimization
Choose between instant inference, fast response, or deep reasoning based on your application needs.
How it works
Every request passes through a 3-tier optimization pipeline.
Semantic Cache Check
85% similarity matching against local buffer. Semantically identical queries return instantly at zero cost.
Intent Analysis
AI classifies your prompt by coding, reasoning, creative, and speed requirements to select the optimal model.
Smart Dispatch
Request routes to the best provider with automatic cascading fallback if primary endpoints fail.
Ready to optimize your AI costs?
Join developers and teams already saving up to 90% on LLM inference costs with intelligent routing.
Features
Everything the routing engine does on every request, plus the platform tooling built around it — wallets, budgets, workspaces, and analytics.
The routing engine
Intelligent Routing
Every prompt is classified by intent — coding, reasoning, creative, or speed-sensitive — then dispatched to the model that best fits that intent at the lowest defensible cost.
Cost Arbitrage (Pay-per-Saving)
We don't mark up inference. We charge a percentage of the difference between what a premium flagship model would have cost and what you actually paid through our routing.
Semantic Caching
85% similarity matching against a local buffer returns semantically identical queries instantly, at zero additional cost.
Cascading Fallbacks
If a primary provider errors or times out, the request automatically falls back across the routing pool instead of failing.
Quality Layer
A quality pass scores every response against the original intent. If it misses the bar, the optimization engine escalates to a higher model and re-runs — then records actual vs predicted quality back into the routing registry.
Logic Layer
IAM-Token extracts the goal, constraints, assumptions, steps, questions, claims, and evidence — a machine-readable map of intent before any token is spent.
Built around it
Managed or BYOK keys
Use our managed keys or bring your own OpenAI / Anthropic / Google keys. Pay-per-Saving still applies at 20% Managed or 50% BYOK.
Wallet, budgets & kill-switches
Prepaid wallet with reserve/settle/refund, auto-recharge, alerts, and a hard kill-switch before a request executes.
Webhooks
HMAC-signed outbound webhooks for usage and billing events.
Workspaces & departments
Seats, roles, shared prompts, per-department cost tracking. SSO planned — not live yet.
Lifecycle analytics
Replay any request: segmentation, intent, router decision, execution, quality scoring.
Security & auth
TOTP MFA, encrypted keys at rest, admin audit logging.
Providers
53 models across 18 brands, one routing engine. We don't lock you into a single vendor's roadmap — the router picks whichever model is actually the right fit, and falls back if a provider errors.
OpenAI
GPT-4o, GPT-4 Turbo, o4-mini, GPT-3.5 Turbo
Anthropic
Claude Opus 4, Claude Sonnet 4, Claude 3.5 Haiku
Gemini 2.5 Pro, Gemini 2.5 Flash, Gemini 2.0 Flash
DeepSeek
DeepSeek R1, DeepSeek V3
xAI
Grok family
Groq
Ultra-low-latency inference on open models
Mistral
Mistral Large, Mistral Small
Qwen
Qwen large-context and coding models
Meta
Llama 3.3 70B, Llama 3.1 8B
Perplexity
Search-grounded generation
Cohere
Command R+, Command R
MiniMax
MiniMax reasoning and chat models
Amazon
Nova model family
MoonshotAI
Kimi long-context models
NVIDIA
NIM-hosted open model endpoints
ByteDance
Seed model family
Sakana AI
Evolved/merged model research releases
Z.ai
Chat and reasoning models
Managed keys
We hold the provider keys, handle rate limits and failover, and bill through your I amToken wallet.
Bring your own keys (BYOK)
Connect your own OpenAI, Anthropic, or Google keys — encrypted at rest — and pay providers directly.
Pricing & plans
Simple monthly plans, each with wallet credit included for inference. The routing engine stretches that credit further — you pay pure passthrough only on what you actually use.
Plus
For individuals & power users
- $16 wallet credit included
- Full routing engine + quality layer
- Semantic caching
- Chat + API access
- Community support
Pro
For developers in production
- $79 wallet credit included
- Everything in Plus
- Production API, unlimited keys
- Budget alerts + kill-switch
- Priority support
Business
For teams & organizations
- $399 wallet credit included
- Everything in Pro
- Workspace, roles & departments
- Custom routing rules
- 24/7 priority support + SLA
Enterprise
High volume & custom needs
- Volume-based pricing
- SSO & procurement
- Custom routing rules + SLA
- Dedicated support
Products
One routing engine, four ways to use it — chat in the browser, call it from your own code, roll it out to a team, or watch exactly how it decides.
Routing decisions, live
Example week · single-vendor flagship spend vs. I amToken routed spend. (Chart lives on the real Products page — design your motion over this area if you want.)
Chat
The routing engine, as a conversation.
Ask anything — every message is routed to the model that fits it best, with semantic caching so you never pay twice for the same question.
Developer API
Drop-in routing for your own app.
Generate an iam_sk_ key and call /v1/route from any codebase. Same engine, same Pay-per-Saving billing.
Workspace
One routing engine, a whole team.
Seats, roles, shared prompts, departments, per-department cost tracking. SSO planned — not live yet.
Routing analytics & lifecycle
See why a model was chosen, not just that it was.
Replay any request as a lifecycle run: segmentation, intent, router decision, execution, quality scoring.
About Us
I amToken exists because picking one AI provider and living with its pricing, its outages, and its roadmap is a bad deal for anyone building on top of language models. We built a routing gateway that sits in front of every major provider, decides which model actually fits each request, and only makes money when that decision saves you money.
That's the whole thesis: route each prompt to the best model at the best price, and charge a cut of the savings versus a premium-flagship baseline — not a markup on spend.
What we believe
Provider-neutral by design
We don't get paid more when you use a more expensive model.
We charge for savings, not spend
If routing doesn't save you money on a request, we don't make money on it either.
Nothing routed in a black box
Intent, model choice, and quality score are visible in lifecycle analytics.
Your keys, your choice
Managed keys or BYOK — same routing, caching, and analytics.
Contact Us
Fill out the form below and we'll get back to you. Or email a specific inbox if you already know who you need.
Send a message
Name, email, subject, and message.
SendGeneral & support
support@iam-token.com
Billing
billing@iam-token.com
Privacy
privacy@iam-token.com
Quickstart
1) Start backend and frontend
make run-gateway cd web && npm run dev
2) Use the Chat UI
Register, then open Chat. Settings for BYOK, Keys for API keys, Dashboard for usage, Routing for quality/cost insights.
3) Call the developer API
curl -H "Authorization: Bearer iam_sk_..." \
-H "Content-Type: application/json" \
-d '{"prompt":"Hello"}' \
http://localhost:8000/v1/route