Summary

Agentara is a 24/7 personal assistant designed for local execution, powered by Claude Code and OpenAI Codex. It addresses the need for a private, integrated AI assistant by offering multi-channel messaging, long-term memory, skills, task scheduling, and session management, benefiting users who require continuous AI support and automation without relying on cloud services.

How It Works

Agentara orchestrates interactions with large language models (Claude Code, OpenAI Codex) through managed sessions, featuring full streaming support for responsive communication. It employs a queue-based task dispatcher (Bunqueue) for serial session execution and cross-session concurrency, enabling robust task scheduling and cron job functionality. The system supports multi-channel messaging (e.g., Feishu/Lark) with message streaming and updating capabilities and includes file/image handling. Data persistence is managed via JSONL files for session history and SQLite for databases, all built on a Bun runtime with a Hono API backend and a React frontend.

Quick Start & Requirements

• Primary install / run command: Clone the repo, then run make install followed by make dev to start both backend and frontend in development mode.
• Non-default prerequisites: Bun (latest), Claude Code installed and configured, OpenAI Codex installed and configured.
• Estimated setup time or resource footprint: On first run, Agentara auto-generates configuration and data directories (~/.agentara). The backend runs on http://localhost:1984 and the frontend dev server on http://localhost:8000.
• Links: Repository: https://github.com/magiccube/agentara.git

Highlighted Details

• Agent-powered sessions with Claude Code and OpenAI Codex, including full streaming support.
• Multi-channel messaging support (e.g., Feishu/Lark) with message streaming and updating.
• Extensive built-in skills for tasks like data analysis, image generation, and technical writing.
• RESTful API powered by Hono, with a React-based web dashboard for managing sessions, tasks, and memory.
• Session persistence via JSONL files and SQLite databases for long-term memory.

Maintenance & Community

The README does not provide specific details on notable contributors, sponsorships, or community channels like Discord/Slack. Contributions are welcomed via standard open-source practices (fork, branch, PR).

Licensing & Compatibility

• License type: MIT License.
• Compatibility notes: The MIT license is permissive and generally compatible with commercial use and closed-source linking.

Limitations & Caveats

The README does not explicitly list limitations, alpha status, or known bugs. The setup requires local installation and configuration of both Claude Code and OpenAI Codex, which may represent a significant setup hurdle.