What are Digital Humans
Starting with Halo v2.0, Halo is no longer just a "one-off AI conversation tool" — it is a platform capable of running persistent AI individuals.
These long-running AI individuals are called Digital Humans.
Digital Humans vs. Regular Conversations
| Regular Conversation | AI Digital Human | |
|---|---|---|
| How it runs | You send a message, AI replies | Runs autonomously in the background, 24/7 |
| Context | Cleared when the conversation ends | Long-term memory, persists across days and weeks |
| Trigger | Manual input | Scheduled tasks, event triggers, Webhooks |
| Execution capability | Full Agent capabilities | Full Agent capabilities (identical) |
| Best for | Instant Q&A, one-off tasks | Continuous monitoring, periodic automation |
In a nutshell: a regular conversation is a temp worker you hire; a Digital Human is a full-time employee.
Digital Human Capabilities
AI Digital Human
├── Identity & Memory → Persistent, maintains context across runs
├── Triggers → Scheduled tasks / file changes / Webhooks / page monitoring
├── Execution Tools → MCP servers / AI browser / shell commands
├── Output Channels → System notifications / email / WeCom / Feishu / DingTalk
└── User Configuration → Keywords, URLs, API keys, and other parametersThe Difference Between MCP, Skill, and A2A
These concepts are easy to confuse — here is a clear breakdown:
| Concept | What it is | Analogy |
|---|---|---|
| MCP | Tool interfaces the AI can call (query databases, operate files, call APIs…) | The toolbox in an employee's hands |
| Skill | A packaged, single capability (send email, search, write a report…) | A skill the employee knows |
| A2A (Google) | A network protocol for Agent-to-Agent communication — solving "how to pass messages" | A walkie-talkie between employees |
| AI Digital Human (DHP) | A complete, independently running AI individual with identity, memory, goals, and triggers | The employee themselves |
To put it plainly:
- MCP and Skills are "tools"; Digital Humans are "the people who use those tools"
- A2A addresses communication format; DHP addresses "who this Agent is, what it can do, and how to distribute it to everyone"
- A Digital Human is not a one-off prompt — it has a persistent identity, long-term memory, and the ability to be triggered externally
Typical Scenarios
- Sentiment monitoring: Searches for brand keywords every hour, immediately pushes an alert when anomalies are detected
- Code review: Listens to GitHub PRs, automatically performs code quality checks and leaves comments
- Daily/weekly report assistant: Summarizes work progress on a schedule and sends it by email
- Information subscriptions: Tracks competitor updates and industry news, periodically pushes summaries
- Task delegation: Hand a long-term goal to a Digital Human and let it keep making progress while you do other things
Next Steps
- Digital Human Store: Install ready-made Digital Humans with one click
- Create a Digital Human: Create your own Digital Human using natural language
- DHP Protocol: Developer-oriented — a deep dive into the protocol specification