Use Cases

MyInner.Health uses ANT to power its supplement scanning pipeline. Users photograph a supplement label, and the app runs it through a 4-stage AI pipeline: OCR text extraction, LLM parsing with web search enrichment, data validation, and a vision fallback for damaged labels — all routed through ANT.

• •OCR via Mistral OCR API (ant:ocr bridge) for instant text extraction from label photos
• •Vision + Search pipeline: LLM parses OCR text and searches the web to complete product info
• •Automatic fallback: if OCR fails, vision model analyzes the raw image directly
• •Project presets: 4 saved API call templates (OCR Scan, Parse & Enrich, Enrich Only, Vision Fallback)
• •Cost-optimized: simple enrichment queries use ant:best (auto-downroutes), only complex parsing uses vision models

CyclingWeather combines weather data with AI analysis to give cyclists actionable ride recommendations. ANT routes weather interpretation queries to the fastest available model — real-time conditions need instant responses, so latency matters more than peak quality here.

• •Real-time weather analysis routed through ant:fast for sub-second responses
• •Natural language ride recommendations generated from structured weather data
• •Web search integration for local event and road condition awareness
• •Cost-effective: weather summaries use fast/free models, detailed analysis escalates only when needed

Antbase itself runs on ANT. The Playground, model intelligence pages, blog image generation, free model scanner, and classification pipeline all use ANT routing. Every internal AI call goes through the same routing engine that users access.

• •Playground: every chat, vision, and OCR request in the Playground is routed by ANT
• •Blog hero images: all 38 blog post images generated via ant:image-gen (Gemini Flash Image)
• •Free model scanner: automated discovery of free models across all providers every 6 hours
• •Custom classifier LLM: the routing classifier itself is an AI model that tags every request
• •Model intelligence: scoring and ranking computed from real routing telemetry data

OpenClaw is an open-source terminal AI coding agent. When connected to ANT, it gets intelligent model routing for every coding task — complex refactors go to Claude Opus or GPT-5, quick edits go to fast models, and experimentation mode uses ant:free for zero cost.

• •ant:coder routes to the best code-generation model available (Claude, GPT, DeepSeek Coder)
• •ant:reasoning activates chain-of-thought for architecture decisions and complex debugging
• •ant:free enables zero-cost experimentation — prototype and learn without spending credits
• •Automatic failover: if Anthropic is down, coding requests seamlessly route to OpenAI or open-source models
• •One config change to connect: set base_url and api_key in OpenClaw settings