MagickMind
  1. Get started
  • Documentation
    • Get started
      • Introduction
      • Quick Start
      • First API Call
    • Core Concepts
      • Projects
      • Corpus
      • Mindspaces
      • End-Users
      • Websocket
    • Authentication
      • Overview
      • OAuth Flow
      • API Keys
    • Guides
      • Creating a Corpus
      • Managing End-Users
    • SDK
      • Python
      • Go
  • API Reference
    • v1
      • Auth
        • Login
        • Refresh Token
      • End-Users
        • Create End-User
        • Get End-User by ID
        • Update End-User
        • Query End-Users
        • Delete End-User
      • Project
        • Create Project
        • Get Project by ID
        • Get Project List
        • Update Project
        • Delete Project
      • Mindspaces
        • Create Mindspace
        • Update Mindspace
        • Get Mindspace by ID
        • Get Mindspace List
        • Delete Mindspace
        • Get Mindspace Messages
        • Delete Mindspace
        • Add Mindspace Users by ID
      • Chat
        • Magick Mind Chat
      • Realtime
        • Websocket Subscribe
      • Runtime
        • Get Effective Personality
        • Invalidate Cache
  1. Get started

Introduction

logo.png

What is Magick Mind?#

Magick Mind is an AI platform that enables developers to build intelligent applications with memory, personality, and knowledge retrieval capabilities. Unlike basic LLM APIs, Magick Mind provides a complete infrastructure for creating AI experiences that remember context, maintain consistent personalities, and ground responses in your own data.

Core Concepts#

Projects#

Projects are organizational containers that group related resources. Think of them as workspaces for different applications or use cases. A customer support bot and an internal knowledge assistant would be separate projects.

Mindspaces#

Mindspaces are isolated conversation contexts that maintain history and memory. Each mindspace can:
Store conversation history
Connect to knowledge bases (corpora) for RAG
Support multiple users for group conversations
Belong to a single project

Personas#

Personas define how your AI behaves and communicates. They include:
Traits - Personality characteristics (empathetic, analytical, friendly)
Tones - Communication styles (professional, casual, warm)
Background - Character history that influences responses
Personas can evolve over time based on interactions while respecting defined boundaries.

Corpus#

A corpus is a collection of documents and knowledge artifacts. When connected to a mindspace, the AI can retrieve relevant information to ground its responses (RAG - Retrieval-Augmented Generation).

End Users#

End users represent the people interacting with your AI. Track them with external IDs for analytics, personalization, and multi-tenant architectures.

Key Features#

Memory & Context#

Conversations in mindspaces maintain full history. The AI remembers what was discussed, enabling natural multi-turn conversations without re-explaining context.

RAG (Retrieval-Augmented Generation)#

Connect your documents to mindspaces. The AI automatically retrieves relevant information to answer questions accurately, reducing hallucinations and grounding responses in your data.

Evolving Personalities#

Personas can learn and adapt within defined boundaries. An AI assistant might become more familiar with a user over time while maintaining its core character.

Multi-Model Support#

Access multiple LLM providers (OpenAI, Anthropic) through a unified API. Switch models or use the best model for each task without code changes.

OpenAI Compatibility#

Drop-in replacement for OpenAI's API. Migrate existing applications by changing the base URL and API key.

Use Cases#

Use CaseFeatures Used
Customer Support BotMindspaces + Corpus (FAQ/docs) + Persona
Knowledge AssistantCorpus + RAG for internal documents
AI CompanionPersona with evolving traits + Memory
Multi-Tenant SaaSProjects + End Users + Tenant isolation
Content GenerationOpenAI-compatible API + Model selection

Next
Quick Start
Built with