Deepself

Give your agent many faces.

---

 

Create precise AI personas that sit between your agent and the language model. Powerful, dynamic, plug-and-play identities your agent can spin up and deploy for specific tasks.

Deepself is a psychology engine that renders identity and essence through computational linguistics and matrix math. It is based on proprietary maps that distill cognition into efficient building blocks.

Deepself personas allow LLMs to speak like real humans: complex, layered, internally consistent yet capable of holding contradictions.

A small amount of training data becomes a fully-formed perspective.

Create personas for:
- Embodying perspective
- Domain expertise (legal, creative, executive)
- Conversation modeling

Train over API or in our apps with content samples attributed to or describing the persona:
- Documents
- Conversations
- Imported content (YouTube, WhatsApp, etc)

Use them with:
- AgentSkills (SKILL.md) and our plugin for OpenClaw agents
- our MCP server and OpenAI-compatible inference API
- our conversation circuits to model group discussions

Personas are objective and can be compared, measured, and calibrated. You or your agent select the best face for the job.
                
For OpenClaw agents:

1. Sign up and get your API key

2. Install via:
   - OpenClaw plugin: 
     github.com/deepself/openclaw-plugin
   
   - AgentSkill: 
     clawdhub install deepself

3. Add your API key to config