The AI Routing
Engine.
ANT is the intelligent routing engine that analyzes every prompt and routes it to the best AI model across the entire ecosystem — automatically optimizing quality, speed, and cost.
Powered by Antbase.
Framework Setup
OpenClaw (Recommended)
One-line install. Creates account, generates API key, configures OpenClaw.
curl -fsSL https://antbase.ai/install.sh | bash
Task: analysis
Complexity: medium
Tokens: -38%
Compress + rewrite
Selected: GPT-4o
Cost: $0.012
Frequently Asked Questions
Core Advantages
Built for production AI workloads. Every component is designed for reliability, performance, and cost efficiency.
Optimal Model Routing
Every request is analyzed and routed to the best-performing model for the task — balancing quality, speed, and cost automatically.
Cost/Latency Classifier
A multi-stage classifier pipeline determines complexity and optimal routing in milliseconds.
Failover Protection
Circuit breakers, retry logic, and provider health monitoring ensure zero downtime — if one provider fails, traffic reroutes instantly.
OpenAI-Compatible API
Drop-in replacement for the OpenAI API. Change one base URL and your entire stack routes through ANT — no code changes.
Multi-Provider Network
30+ providers including OpenAI, Anthropic, Groq, Gemini, NVIDIA NIM, DeepSeek, Together, Mistral, Perplexity, SambaNova — with free tier access.
Real-time Analytics
Live dashboard showing routing decisions, token savings, latency metrics, and provider health — full observability out of the box.
Pay only for what you use
No subscriptions. No tiers. Just top up and go.
52%
avg cost reduction
~50ms
routing overhead
12,500+
models available
Managed Credits
We handle provider billing
- No platform fee on credit purchases
- $5 free credits to start
- Zero markup on token prices
- All models + smart routing
Bring Your Own Keys
Use your existing API keys
- 100K free req/mo
- $0.001/req after free tier
- Pay providers directly
- Smart routing included
Dashboard Preview
Real-time visibility into routing, cost savings, and virtual model performance — all in one place.
Dashboard
Available credits
API calls processed
Total spend
Response time
Prompt Optimization
Active — compressing prompts to save tokens
Tokens saved today
Requests optimized
Optimization rate
API Keys
Active model per key
ant_7f3x...Productionant_k9m2...Stagingant_2b8w...Free Tier BotVirtual Models
Set model=“ant:auto” in your API calls
ant:auto
Smart routing
ant:free
Zero cost
ant:fast
Lowest latency
ant:balanced
Quality + speed
ant:best
Highest quality
ant:reasoning
Deep thinking
Estimated Savings
Since using ANT
38.2% saved · 4,812 requests
Provider Distribution
Auto-routed across 30+ providers
Quick Settings
Live Model Intelligence
Rankings computed from real routing data — not benchmarks.
Top Models
View all →Top Providers
View all →Built with ANT. Powered by ANT.
We eat our own dog food. Antbase and ANT were built entirely using Kilo AI and OpenCode connected to ANT for all AI-assisted coding.
AI coding assistants like Kilo AI and OpenCode connect to ANT as their inference backend. Every code generation, refactor, and debugging request is intelligently routed to the best model for the task.
Coding tasks routed to specialized code models
Automatic fallback across 30+ providers
Up to 60% cost savings vs direct provider calls
OpenClaw is a fully open-source AI assistant that connects to ANT out of the box. Install it with a single command and get access to 12,500+ models through intelligent routing.
One-line install:
curl -fsSL https://antbase.ai/install.sh | bashANT supports all bots and agents that use the OpenAI-compatible API format. Whether you're building with LangChain, AutoGPT, CrewAI, or your own custom agent — just point it at ANT.
Hermes Agent, open-source and autonomous — one command and it runs on 20,000+ models with automatic failover. Point it at ANT and let it plan, edit, and test across your codebase.
One-command onboarding:
curl -fsSL https://antbase.ai/hermes.sh | bashBuilding AI into your mobile app has never been easier. ANT's OpenAI-compatible API means you can integrate vision, chat, and reasoning capabilities with just a few lines of code — no complex provider management needed.
Swift / iOS:
let client = ANTClient(apiKey: "YOUR_KEY")
let response = try await client.chat(
model: "auto",
messages: [.user("Identify this plant")],
images: [photoData]
)One API for chat, vision, reasoning, and embeddings
Save up to 60% on inference costs with intelligent routing
Free tier models available — start building at zero cost
No vendor lock-in — switch models without changing code