agentic-in/elephant-agent

What it solves

Elephant Agent addresses the limitation of task-oriented AI by shifting the focus from executing isolated tasks to understanding the person using the AI. It aims to create a persistent, evolving understanding of a user's identity, environment, and goals to help shape long-term life and work trajectories rather than just providing quick answers.

How it works

The system builds a Personal Model consisting of four lenses: Identity (values and preferences), World (people and projects), Pulse (current priorities and energy), and Journey (lessons and growth patterns). This model is developed through interaction, correction, and questioning.

Once established, the agent helps design Paths—long-running directions like health, research, or coding—which it breaks down into actionable steps. It can deploy "baby elephants" (sub-agents) for specific tasks and returns to the user at critical checkpoints for judgment.

Who it’s for

Individuals seeking a personal AI companion that grows with them over time, manages long-term goals (Paths), and integrates with their daily digital workspace (macOS, CLI, or Cloud).

Highlights

• Personal Model Map: A visual and inspectable representation of what the AI understands about the user.
• Path Management: Ability to turn long-term goals into structured steps and checkpoints.
• Multi-Surface Access: Available as a native macOS app for a full workspace experience or a CLI + Dashboard for Linux/Cloud users.
• Extensible Tooling: Integration with browser, filesystem, MCP, and various messaging platforms (e.g., Discord, WeChat).
• Flexible Model Providers: Supports both local and hosted LLM providers.