eggent-ai/eggent

What it solves

Eggent is a local-first AI workspace designed to organize AI agents into project-scoped environments. It solves the problem of fragmented AI interactions by providing a centralized orchestration layer where agents have persistent memory, project-specific instructions, and access to a dedicated set of files and tools.

How it works

Eggent operates as a browser-based dashboard and agent runtime. It uses a directory-backed configuration system where each project is a folder containing its own context, memory, and model settings. An "Orchestrator" agent manages the workspace when no specific project is selected. The system integrates with various LLMs via provider keys and supports the Model Context Protocol (MCP) and custom skills. It also includes a local-first storage approach where all application data is stored in a data/ directory for easy backup.

Who it’s for

It is intended for users who need a structured way to manage multiple AI agents across different projects, developers who want to integrate AI agents into external services via an HTTP API, and users who want a private, self-hosted AI workspace with local voice transcription.

Highlights

• Project-scoped Agents: Each project has its own instructions, memory, and MCP configuration.
• Pipelines: Ability to run multiple agents in a sequence with artifact handoff between steps.
• Local-first Storage: All app data is stored as files in a local directory.
• External API: A bearer-token protected HTTP API to connect Eggent to bots, CRMs, or other backend services.
• Telegram Integration: A built-in bot that supports formatted responses and voice message transcription.
• Local Voice Input: Uses whisper.cpp for local speech-to-text transcription without external APIs.
• MCP & Skills Support: Support for the Model Context Protocol and project-local skills.