# Introduction

> You're reading the Mindset v3 documentation. v3 is no longer sold; this is reference for existing customers. Mindset 4 docs are at /docs.

Mindset API & SDK Overview

## Introduction

The Mindset API and SDK provide a powerful platform for developers who want to build AI-powered applications without the complexity of managing AI infrastructure.

**The API** gives you programmatic access to upload content, manage knowledge contexts, configure agents, and track analytics - everything you need to power intelligent applications.

**The SDK** provides pre-built JS components and utilities that let you embed multi-media, conversational agents into your application with just a few lines of code - no need to build chat interfaces from scratch.

> **Tip:**
>
> ### So why do you need this - why not just work with one of the agent frameworks or the LLM directly?
>
> Of course, that is an option, but the Mindset API and SDK provide a more streamlined and efficient way to work, where much of the heavy lifting is already done for you.
>
> This allows you to focus on building your application and delivering value to your users, rather than worrying about the plumbing.

## "But agents are just prompts, data, and configuration, right?"

Yes, at the surface level, an agent might look like a fancy chatbot with some prompts and a knowledge base. But that's like saying a car is just an engine, wheels, and a steering wheel. Technically true, but you're missing about 99% of what makes it actually work in the real world.

Here's what you're *really* building when you create production-ready agents:

- **Context Management**: How do you maintain conversation state across sessions? What happens when users switch topics mid-conversation? How do you handle context windows that overflow?
- **Retrieval Engineering**: Sure, you can throw documents at a vector database, but how do you handle conflicting information? How do you ensure the agent retrieves the *right* context, not just *similar* context?
- **Safety & Guardrails**: What happens when your agent hallucinates? How do you prevent prompt injection? What about when users try to jailbreak your carefully crafted prompts?
- **Performance at Scale**: Your prototype works great with 10 users. What about 10,000? How do you handle rate limiting, caching, and cost optimization across multiple LLM providers?
- **Tool Orchestration**: Agents that can only chat are toys. Real agents need to trigger actions, call APIs, and orchestrate complex workflows. How do you handle tool failures, retries, and partial completions?
- **Observability**: When your agent gives a wrong answer, how do you trace it back through the retrieval pipeline, prompt construction, and model reasoning to figure out what went wrong?

The difference between a weekend hackathon agent and a production system is the same as the difference between a paper airplane and a Boeing 747. They both fly, but only one will get you where you need to go reliably.

## Do you really want to build the following yourself?

- **Content Management**: Uploading, managing, and organizing content efficiently and making it available to agents ensuring the right users see only the right content.
- **Knowledge Contexts**: Ingest, organize, and manage content libraries for agents, including integration with external sources.
- **Label Management**: Creating and managing labels for content classification and ensuring that every agent interaction is properly tracked and categorized.
- **Analytics & Reporting**: Track agent performance, usage, and user interactions with built-in BI dashboards and reporting tools.
- **Testing Environment**: Comprehensive testing framework with the ability to manage agent test configurations across different LLMs and compare performance results.
- **Tool Integration**: Enable agents to use tools (like clarification or URL injection) and manage tool invocation logic automatically.
- **Policy & Guardrails**: Configure agent policies, guardrails, and safeguards to control agent behavior and ensure compliance.
- **Bias Analysis**: Automatically analyze agent knowledge for unexpected or risky concepts and trace them to their source.
- **Multi-Tenant Management**: Support for multiple tenants, each with their own agents, knowledge, and analytics.
- **Human/User Management**: Securely manage users (humans) and accounts with advanced authentication and privacy controls.
- **Integrations**: Out-of-the-box integrations with 500+ apps and services via a low-code GUI and embedded iPaaS.

## Surely you would rather focus on the good stuff

- **Building Your Core Product**: Focus on what makes your application unique instead of rebuilding AI infrastructure from scratch.
- **Seamless Agent Integration**: Embed intelligent agents into your application with just a few lines of code, complete with customizable UI components.
- **Rapid Prototyping & Deployment**: Go from concept to production in days, not months, with pre-built components and workflows.
- **Leveraging 500+ Integrations**: Connect to CRMs, payment systems, communication channels, and content sources through our embedded iPaaS without building custom connectors.
- **Creating Exceptional User Experiences**: Design engaging, conversational interfaces that your users will love, backed by enterprise-grade AI capabilities.
- **Building for the Future of Interfaces**: Prepare for a world where software is consumed through minimal interfaces - voice commands, ambient computing, and eventually just an earpiece.
- **Voice-First Applications**: Develop applications that work seamlessly with voice interactions, reducing the dependency on traditional screens and keyboards.
- **Ambient Intelligence**: Create software that integrates naturally into users' environments, anticipating needs and responding through the most natural interface possible - conversation.
- **Scalable Architecture**: Build applications that grow with your business, supported by our multi-tenant, cloud-native infrastructure.
- **Innovation Over Infrastructure**: Spend your time solving customer problems and creating value, not managing servers, databases, and AI model configurations.

## The Bottom Line

While you *could* spend the next 6-12 months building your own AI infrastructure, wrestling with vector databases, prompt engineering at scale, and debugging mysterious agent behaviors at 3 AM, we've already solved these problems.

The Mindset platform handles the complex, boring stuff so you can focus on what you're actually good at - building amazing products that users love.

Ready to get started? Let's dive into the API and SDK documentation.