Naagmani is an enterprise-grade AI Operating System & Gateway Runtime designed to govern, optimize, and secure all Large Language Model (LLM) traffic across your engineering ecosystem.

Often described as the "Android OS for AI", Naagmani acts as a centralized intelligent runtime layer between your application code and the fragmented landscape of foundational model providers (OpenAI, Anthropic Claude, Google Gemini, DeepSeek, and self-hosted open-source models).

Rendering diagram...

graph TD
    subgraph Clients["Your Client Applications"]
        WebApp["Web & Mobile Apps"]
        Backend["Backend Microservices"]
        Agents["Autonomous AI Agents"]
    end

    subgraph NaagmaniOS["Naagmani AI Operating System"]
        Auth["BYOK Key Vault & Scoped Tokens"]
        Router["Smart Model Router & Failover"]
        Governance["AI Guardrails, DLP & Policies"]
        Telemetry["Provider Attempt Accounting & FinOps"]
        MCP["MCP Gateway & Tool Registry"]
    end

    subgraph Upstream["AI Providers & Execution"]
        OpenAI["OpenAI (GPT-4o, o3-mini)"]
        Anthropic["Anthropic (Claude 3.5 Sonnet)"]
        Google["Google Gemini (2.5 Flash, Pro)"]
        DeepSeek["DeepSeek (V3, R1)"]
        SelfHosted["Self-Hosted (vLLM, Ollama)"]
    end

    Clients --> NaagmaniOS
    NaagmaniOS --> Upstream

Why Do You Need an AI Operating System? #

Building modern AI applications directly against individual provider APIs introduces critical enterprise bottlenecks:

Challenge with Direct Provider Calls	How Naagmani Solves It
Provider Lock-In & Breaking Changes	Unified OpenAI-compatible API format across all providers. Switch models with zero code changes.
Outages & Rate Limit Disruptions	Automatic retry cascades and dynamic failover across providers in under 50ms.
Sprawling API Keys & Security Risks	Centralized Bring-Your-Own-Key (BYOK) encrypted vault and scoped Project Service Tokens.
Runaway Costs & Unpredictable Bills	Hierarchical FinOps spending caps (Organization $\rightarrow$ Project $\rightarrow$ Member) with real-time token tracking.
Tool Calling & Agent Fragmentation	Native Model Context Protocol (MCP) server integration and tool execution sandboxing.

The Naagmani Mental Model #

To understand how Naagmani organizes workloads, consider the following structural hierarchy:

Organization: The top-level account and billing tenant (e.g. Acme Corp). Governs aggregate spending limits, team members, and enterprise audit logs.
Project: An isolated application or business initiative (e.g. Customer Support Bot, Internal Copilot).
Environment: Deployment stages within a project (Test and Production) with segregated secrets, routing rules, and rate limits.
Service Tokens & API Keys: Scoped credentials issued to services or developers with explicit capability matrices and optional member attribution.
Smart Routing & Attempts: Intelligent multi-hop dispatch engine that executes upstream provider calls, logs durable latency and token accounting, and handles failovers automatically.

Next Steps #

Explore the business and engineering rationale: Why Naagmani?
Learn about the multi-plane system architecture: Architecture Overview
Send your first AI request in under 3 minutes: Quickstart Guide
Access the web management console: Developer Portal Overview

What is Naagmani?

Why Do You Need an AI Operating System? #

The Naagmani Mental Model #

Next Steps #