# Emotion Machine

> AI character infrastructure with persistent memory, behavioral state, and full-duplex voice. San Francisco.

## Products

- [Personality Machine](https://www.emotionmachine.com/personality-machine): Developer platform for AI characters that maintain identity across conversations and build relationships through persistent memory, profiles, behaviors, and context engineering. REST API, visual builder, and conversation dashboard.
- [Dialog Machine](https://www.emotionmachine.com/dialog-machine): AI phone agent that places real calls, navigates conversations with humans, and returns structured data and transcripts. 1,000+ calls completed.

## Technical Essays

- [How Memory Works at Emotion Machine](https://www.emotionmachine.com/blog/how-memory-works): Deep walkthrough of the memory system — vector-based semantic search, LLM-managed scratchpad, importance scoring, and retrieval strategies.
- [How to Build Realistic AI Companions](https://www.emotionmachine.com/blog/realistic-ai-companions): Techniques for building AI that feels consistent and relational — conversation phases, dialogue patterns, memory, proactive engagement, active listening, and sense of time.
- [AI Companions Are the Next Interface](https://www.emotionmachine.com/blog/ai-companions-are-the-next-interface): Why conversational AI companions are becoming the primary interface for how people interact with technology.
- [Adding Open-Ended Conversations to Your Products](https://www.emotionmachine.com/blog/adding-open-ended-conversations): How to integrate open-ended conversational AI into existing products without rebuilding from scratch.

## API

- [API Reference](https://www.emotionmachine.com/API_V2_REFERENCE.md): Full REST API documentation — companions, relationships, messages (send, stream, connect, voice), memory, profiles, behaviors, sessions, knowledge, tools, inbox.

## Key Concepts

- **Companions**: AI characters with defined personality, voice, boundaries, and knowledge.
- **Relationships**: Persistent user-character pairs that accumulate memory, profile data, and behavioral state across conversations.
- **Memory**: Structured facts scored by importance and retrieved by relevance. Two modes: vector search and LLM-managed scratchpad.
- **Behaviors**: Event-driven functions that fire on keywords, schedules, idle time, or API calls. Give characters initiative and reactivity.
- **Context Engineering**: Eight-layer prompt assembly — system prompt, behavior injections, memory, knowledge, profile, session state, history, user message. Configurable per companion, relationship, or turn.
- **Voice**: Full-duplex speech-to-speech with natural interruptions, backchannels, and custom voice via voice prompt embedding.

## Contact

- Website: https://www.emotionmachine.com
- Email: hello@emotionmachine.ai
- Founded by Ege Ozgirin and Emre Sarbak. Backed by Gradient Ventures. Based in San Francisco.