aeonfun/aeon

The most autonomous agent framework. No approval loops. No babysitting. Configure once, forget forever.

What it solves

Aeon is an autonomous agent framework designed to handle recurring, unattended work that typically requires constant human supervision. It eliminates "approval loops" and "babysitting" by allowing users to configure agents to perform complex tasks—such as shipping code, conducting security research, and monitoring markets—on a schedule without needing manual confirmation for every action.

How it works

Aeon operates using "skills," which are simple Markdown files containing frontmatter (configuration) and a prompt. These skills are executed based on cron schedules defined in a configuration file (aeon.yml).

Key operational features include:

  • Self-Healing Loop: Every run is scored by an LLM (Haiku). If a skill fails or its quality drops, a sequence of internal skills (heartbeat, skill-health, skill-repair, and self-improve) automatically diagnoses and patches the prompt or configuration.
  • Fleet Management: Aeon can spawn specialized copies of itself (instances) to handle specific domains (e.g., a dedicated crypto-tracker) and coordinate them via a fleet control system.
  • Harnesses: It supports multiple LLM providers and gateways (Anthropic, OpenRouter, Grok, etc.) to power the agents.
  • Integration: It uses the GitHub CLI (gh) for repository management and supports various notification channels (Telegram, Discord, Slack, Email) to alert users of results.

Who it’s for

Aeon is built for developers, security researchers, and power users who want to automate high-stakes, recurring workflows—like vulnerability scanning, feature implementation, and deep research—without having to manually trigger or approve every step.

Highlights

  • Autonomous Code Shipping: Can implement features and deploy live web apps to Vercel unprompted.
  • Automated Security Research: Includes a vuln-scanner that finds and responsibly discloses real vulnerabilities in open-source projects.
  • Self-Evolving Skills: Can generate new skills from a single sentence or evolve existing ones through scored variations.
  • Zero-Install Setup: Provides an "Aeon setup skill" that allows users to configure and debug their instance using plain language via agents like Claude Code.
  • Built-in Guardrails: Includes read-only modes, circuit-breakers for financial transactions, and an optional authorization layer to prevent runaway autonomy.