Background intelligence that evolves with you.
Autonomous agent on GitHub Actions, powered by Claude Code. 51 skills across research, dev tooling, crypto monitoring, and productivity — all off by default, enable what you need.

OpenClaw is great if you need real-time responses and have infra to run it on. Aeon is for everything else:

• Cheaper — runs on GitHub Actions, free for public repos, ~$2/mo otherwise. No server.
• Built for background tasks — digests, monitoring, research, writing. You don't need sub-second latency for any of that.
• Doesn't break — no daemon to crash, no process to restart. If GitHub Actions is up, Aeon is up. Failed skill? Next cron tick retries it.
• 5-minute setup — fork, add secrets, flip skills on. No Docker, no self-hosting, no config files beyond one YAML.

git clone https://github.com/aaronjmars/aeon
cd aeon && ./aeon

Click on http://localhost:5555 to open the dashboard in your browser. From there:

1. Authenticate — add your Claude API key or OAuth token
2. Add a channel — set up Telegram, Discord, or Slack so Aeon can talk to you (and you can talk back)
3. Pick skills — toggle on what you want, set a schedule, and optionally set a var to focus each skill
4. Push — one click commits and pushes your config to GitHub, Actions takes it from there

You can also schedule and trigger skills by messaging Aeon directly on Telegram — just tell it what you want.

Set one of these — not both:

Getting an OAuth token:

claude setup-token   # opens browser → prints sk-ant-oat01-... (valid 1 year)

Aeon publishes articles to a browsable gallery via GitHub Pages. After merging, enable it in Settings → Pages → source Deploy from a branch, branch main, folder /docs.

The gallery will be live at https://<username>.github.io/aeon.

The update-gallery skill syncs articles/*.md → docs/_posts/ with proper Jekyll frontmatter on a weekly schedule. Enable it in aeon.yml to keep the gallery up to date automatically.

Subscribe to Aeon's article output via Atom feed:

https://raw.githubusercontent.com/<owner>/<repo>/main/articles/feed.xml

Add this URL to any RSS reader (Feedly, Miniflux, NetNewsWire, etc.) to get new articles as they're published. The feed is regenerated daily or after each content-generating skill runs.

The only skill enabled by default. Runs every 3 hours as a fallback catch-all.

Every run: reads recent memory and logs, checks for stalled PRs (>24h), flagged memory items, urgent issues, and skills that haven't run on schedule. Deduplicates against the last 48h of logs — it won't re-notify you about something it already flagged.

Nothing to report → logs HEARTBEAT_OK, exits, no commit. Something needs attention → sends a concise notification.

Heartbeat is listed last in aeon.yml so it only runs when no other skill claims the slot.

All scheduling lives in aeon.yml:

skills:
  article:
    enabled: true               # flip to activate
    schedule: "0 8 * * *"       # daily at 8am UTC
  digest:
    enabled: true
    schedule: "0 14 * * *"
    var: "solana"               # topic for this skill

Standard cron format. All times UTC. Supports *, */N, exact values, comma lists.

Order matters — the scheduler picks the first matching skill. Put day-specific skills (e.g. Monday-only) before daily ones. Heartbeat goes last.

Every skill accepts a single var — a universal input that each skill interprets in its own way:

If var is empty, each skill falls back to its default behavior (scan everything, auto-pick topics, etc.). Set it from the dashboard or pass it when triggering manually.

The default model for all skills is set in aeon.yml:

model: claude-opus-4-6

You can change it from the dashboard header dropdown. Options: claude-opus-4-6, claude-sonnet-4-6, claude-haiku-4-5-20251001. Per-run overrides are also available via workflow dispatch.

Edit .github/workflows/messages.yml:

schedule:
  - cron: '*/5 * * * *'    # every 5 min (default)
  - cron: '*/15 * * * *'   # every 15 min (saves Actions minutes)
  - cron: '0 * * * *'      # hourly (most conservative)

Claude only installs and runs when a skill actually matches.

To reduce usage: switch to */15 or hourly cron, disable unused skills, keep the repo public.

Set the secret → channel activates. No code changes needed.

Telegram: Create a bot with @BotFather → get token + chat ID.
Discord: Outbound: Channel → Integrations → Webhooks → Create. Inbound: discord.com/developers → bot → add channels:history scope → copy token + channel ID.
Slack: api.slack.com → Create App → Incoming Webhooks → install → copy URL. Inbound: add channels:history, reactions:write scopes → copy bot token + channel ID.

Default polling has up to 5-min delay. Deploy this ~20-line Cloudflare Worker for ~1s response:

export default {
  async fetch(request, env) {
    const { message } = await request.json();
    if (!message?.text || String(message.chat.id) !== env.TELEGRAM_CHAT_ID)
      return new Response("ignored");

    // Advance offset so polling doesn't reprocess
    await fetch(`https://api.telegram.org/bot${env.TELEGRAM_BOT_TOKEN}/getUpdates?offset=${message.update_id + 1}`);

    // Trigger GitHub Actions immediately
    await fetch(`https://api.github.com/repos/${env.GITHUB_REPO}/dispatches`, {
      method: "POST",
      headers: { Authorization: `token ${env.GITHUB_TOKEN}`, Accept: "application/vnd.github.v3+json" },
      body: JSON.stringify({ event_type: "telegram-message", client_payload: { message: message.text } }),
    });

    return new Response("ok");
  },
};

Set env vars (TELEGRAM_BOT_TOKEN, TELEGRAM_CHAT_ID, GITHUB_REPO, GITHUB_TOKEN) in Cloudflare dashboard, then point your bot's webhook at the worker URL.

The built-in GITHUB_TOKEN is scoped to this repo only. For github-monitor, pr-review, issue-triage, and feature to work on your other repos, add a GH_GLOBAL personal access token.

Setup: github.com/settings/tokens → Fine-grained → set repo access → grant Contents, Pull requests, Issues (all read/write) → add as GH_GLOBAL secret.

Skills use GH_GLOBAL when available, fall back to GITHUB_TOKEN automatically.

./add-skill BankrBot/skills --list          # browse a repo's skills
./add-skill BankrBot/skills bankr hydrex   # install specific skills
./add-skill BankrBot/skills --all           # install everything

Or discover programmatically:

npx skills find "crypto trading"

Installed skills land in skills/ and are added to aeon.yml disabled. Flip enabled: true to activate.

Every skill in this repo is independently installable. Browse the full catalog in skills.json or use the CLI:

# Install a single skill
./add-skill aaronjmars/aeon token-alert

# Install multiple skills
./add-skill aaronjmars/aeon token-alert polymarket hn-digest

# Install everything
./add-skill aaronjmars/aeon --all

# Browse available skills
./add-skill aaronjmars/aeon --list

Package any skill as a standalone directory (with a generated README) for sharing:

./export-skill token-alert              # exports to ./exports/token-alert/
./export-skill token-alert --tar        # also creates a .tar.gz archive
./export-skill --list                   # list all exportable skills

Label any GitHub issue ai-build → workflow fires → Claude reads the issue, implements it, opens a PR.

Don't need the full agent? Grab individual workflow templates and drop them into any repo. These work with GitHub Agentic Workflows — plain Markdown files in .github/workflows/ that run with any supported agent engine.

# Copy a template into your repo
curl -O https://raw.githubusercontent.com/aaronjmars/aeon/main/workflows/pr-review.md
mv pr-review.md .github/workflows/
git add .github/workflows/pr-review.md && git commit -m "Add PR review workflow"

Each template is self-contained Markdown — edit the trigger, criteria, or format to fit your project. See workflows/README.md for details.

By default Aeon has no personality. To make it write and respond like you, add a soul:

1. Fork soul.md and fill in your files:
   • SOUL.md — identity, worldview, opinions, interests
   • STYLE.md — voice, sentence patterns, vocabulary, tone
   • examples/good-outputs.md — 10–20 calibration samples
2. Copy into your Aeon repo under soul/
3. Add to the top of CLAUDE.md:

## Identity

Read and internalize before every task:
- `soul/SOUL.md` — identity and worldview
- `soul/STYLE.md` — voice and communication patterns
- `soul/examples.md` — calibration examples

Embody this identity in all output. Never hedge with "as an AI."

Every skill reads CLAUDE.md, so identity propagates automatically.

Quality check: soul files work when they're specific enough to be wrong. "I think most AI safety discourse is galaxy-brained cope" is useful. "I have nuanced views on AI safety" is not.

CLAUDE.md                ← agent identity (auto-loaded by Claude Code)
aeon.yml                 ← skill schedules + enabled flags
skills.json              ← machine-readable skill catalog (51 skills)
./notify                 ← multi-channel notifications
./add-skill              ← import skills from GitHub repos
./export-skill           ← package skills for standalone distribution
./generate-skills-json   ← regenerate skills.json from SKILL.md files
docs/                    ← GitHub Pages gallery (Jekyll)
soul/                    ← optional identity files
skills/                  ← each skill is SKILL.md (Agent Skills format)
  article/
  digest/
  heartbeat/
  ...                    ← 51 skills total
workflows/               ← GitHub Agentic Workflow templates (.md)
memory/
  MEMORY.md              ← goals, active topics, pointers
  topics/                ← detailed notes by topic
  logs/                  ← daily activity logs (YYYY-MM-DD.md)
dashboard/               ← local web UI
.github/workflows/
  aeon.yml               ← skill runner (workflow_dispatch, issues)
  messages.yml           ← message polling + scheduler (cron)

This repo is a public template. Run your own instance as a private fork so memory, articles, and API keys stay private.

# Pull template updates into your private fork
git remote add upstream https://github.com/aaronjmars/aeon.git
git fetch upstream
git merge upstream/main --no-edit

Your memory/, articles/, and personal config won't conflict — they're in files that don't exist in the template.

Messages not being picked up?

GitHub has two requirements for scheduled workflows:

1. The workflow file must be on the default branch — crons on feature branches don't fire.
2. The repo must have recent activity — GitHub disables crons on repos with no commits in 60 days. New template forks need one manual trigger to activate.

Fix: Actions → Messages → Run workflow (manual trigger). After that, the cron activates automatically.

Support the project : 0xbf8e8f0e8866a7052f948c16508644347c57aba3