Overview
Open Source OOP Framework for AI Agents — Class is an Agent, Agent is an API
Welcome to OpenLiate
OpenLiate is the sovereign, open-source OOP framework for autonomous AI agents. Built on the LIATE mathematical decomposition rule, it unifies reasoning loops, persistent memory, and Model Context Protocol (MCP) tool execution into production-ready API microservices.
npx openliate run agent.jsonSelected for Sarvam AI Startup Program 2026 🇮🇳
OpenLiate natively supports sovereign Indic models including Sarvam 105B and Bulbul voice across 22+ Indian languages.
The LIATE Mathematical Foundation
In calculus, the LIATE Rule is the universal mnemonic used to break down complex mathematical integration by parts:
"Don't build AI agents. Integrate by parts."
| Calculus LIATE Rule | OpenLiate LIATE Architecture | Purpose |
|---|---|---|
L — Logarithmic | [L] — Large Models Layer | Sarvam AI 105B/2B, local Ollama, cloud LLMs |
I — Inverse Trig | [I] — Integration & Memory | Sessions, Postgres, state & persistent memory |
A — Algebraic | [A] — Agent Harness Layer | Autonomous ReAct reasoning loop & step guardrails |
T — Trigonometric | [T] — Tools & MCP Layer | Model Context Protocol v2 (stdio & HTTP servers) |
E — Exponential | [E] — Environment & Sandbox | Process isolation, deterministic guardrails |