# aeon

> The most autonomous agent framework. Open source, MIT licensed. Give it a
> direction and it leverages 86 skills to get the work done - running
> unattended on GitHub Actions with self-healing loops, reactive triggers, and
> persistent memory. Not affiliated with AEON, the agentic-economy payments
> network, or aeon-toolkit, the Python time-series ML library.

You own everything: your repo, your keys, your GitHub Actions, your memory.
aeon is a set of GitHub Actions workflows plus a Next.js dashboard; Aeon
Connect (https://www.aeon.fun/connect) hosts that dashboard and a remote MCP server at
https://www.aeon.fun/connect/mcp that drives your own instance from ChatGPT, Claude or
any MCP client. There is no hosted runtime - runs happen on the operator's own
GitHub Actions.
Two files define an
agent: SOUL.md (voice) and STRATEGY.md (intent - one metric, a few
priorities, hard limits). Every skill reads both before it acts; each run is
scored 1-5 by Claude Haiku, and skills that fail repeatedly are diagnosed and
patched automatically.

## When to use aeon

Reach for aeon for recurring, unattended work an operator does not want to
babysit:

- Scheduled briefs, digests, and changelogs
- Continuous monitoring (markets, onchain activity, GitHub repos, mentions)
- Autonomous code work (PR reviews, vulnerability scanning and disclosure)
- Social and BD automation (drafting posts, replies, outreach radar)
- Self-evolving fleets (writes new skills, forks into specialised repos)

## Get started

Three ways to stand up your own aeon instance, fastest first:

1. Browser (recommended): [Aeon Connect](https://www.aeon.fun/connect). Sign in with
   GitHub, create your aeon repo (a public fork or a private copy of
   aeonfun/aeon), install the Aeon Connect GitHub App on that one repo, connect
   a model (a Claude subscription token, an API key, or another login), and
   pick skills. No clone, no terminal.
2. Coding agent: paste one line into Claude Code, Codex, Hermes or OpenClaw:
   "read https://www.aeon.fun/skills/aeon.md and follow the instructions to set up your
   aeon agent". The hosted skill sends the user to Aeon Connect, or runs the
   terminal setup for them.
3. Terminal: `git clone https://github.com/aeonfun/aeon && cd aeon && ./aeon init` (needs
   Node.js 20+ and the GitHub CLI). `./aeon` opens the same dashboard locally.

## Developer & integration resources (machine-readable)

- [Aeon Connect](https://www.aeon.fun/connect): the hosted aeon dashboard - sign in with GitHub to create, configure and run an instance in the browser
- [Developer resources](https://www.aeon.fun/developers): API, MCP server, OpenAPI, CLI, auth, rate limits, versioning/deprecation policy
- [Aeon Connect MCP](https://www.aeon.fun/docs#mcp-hosted): drive your own agent from ChatGPT, Claude, Claude Code or Cursor - list and run skills, follow runs, read output and memory, toggle skills, report problems (send_feedback). Remote MCP over Streamable HTTP at https://www.aeon.fun/connect/mcp with OAuth (GitHub sign-in, then pick your agent repo). Claude Code: `claude mcp add --transport http aeon https://www.aeon.fun/connect/mcp`
- [auth.md](https://www.aeon.fun/auth.md): how an agent gets an OAuth token for the Aeon Connect MCP (discovery, client registration, PKCE, refresh, revocation)
- [aeon docs MCP](https://www.aeon.fun/mcp): public, read-only MCP server over Streamable HTTP for searching aeon's docs and skills. Server card: [/.well-known/mcp.json](https://www.aeon.fun/.well-known/mcp.json)
- [OpenAPI spec](https://www.aeon.fun/openapi.json): every reachable endpoint (/ask, /mcp, /api/feedback, discovery docs), typed for function calling
- [NLWeb /ask](https://www.aeon.fun/ask): POST a natural-language question for ranked schema.org results (`prefer: streaming=true` for SSE)
- [Feedback](https://www.aeon.fun/api/feedback): found a broken page or route, a wrong doc, something missing, or have an idea? POST JSON `{ "kind": "bug|error|idea|missing", "title": "...", "details": "...", "page": "/docs" }` to https://www.aeon.fun/api/feedback. Content-Type must be application/json. No key needed, 10 reports per hour. 201/200 `{ "id": ... }` = sent; 400/415/429 = do not retry; 502/503 = skip. Aaron reads every report
- [/agents.md](https://www.aeon.fun/agents.md): agent guide with the live skill catalog
- [/index.md](https://www.aeon.fun/index.md): this page as markdown (append .md, or send Accept: text/markdown)
- [/docs.md](https://www.aeon.fun/docs.md): the full /docs guide as markdown (or send Accept: text/markdown to /docs)
- [/llms-full.txt](https://www.aeon.fun/llms-full.txt): this summary plus the whole docs guide in one file
- [Agent Skills index](https://www.aeon.fun/.well-known/agent-skills/index.json): discovery index with a content digest
- [API catalog](https://www.aeon.fun/.well-known/api-catalog): RFC 9727 linkset of all of the above
- [ai-catalog](https://www.aeon.fun/.well-known/ai-catalog.json): ARD manifest naming every agentic resource

API conventions (full policy at https://www.aeon.fun/developers): v1, URL-path versioned (`https://www.aeon.fun/v1/ask`) with an `API-Version` header, standard RateLimit headers, and `Deprecation` (RFC 9745) + `Sunset` (RFC 8594) headers 90+ days ahead of any change.

CLI: an aeon instance is driven by its own `./aeon` CLI (in its repo) - authenticate, pick skills, schedule. Also ships as an installable Claude Code plugin from `plugin/` in the source repo.

## Documentation

- [Docs & skill catalog](https://www.aeon.fun/docs): every skill and its schedule
- [Skills](https://www.aeon.fun/skills): one page per skill (what it does, its input, keys, schedule, install line) at https://www.aeon.fun/skills/<slug>
- [Architecture](https://www.aeon.fun/architecture): run loop, self-healing, chains, fleet spawning
- [Security](https://www.aeon.fun/security): vulnerabilities aeon has found and fixed
- [Security advisory: Omarchy: code execution from an installed third-party theme](https://www.aeon.fun/security/advisories/omarchy-theme-install-code-execution): GHSA-mxm5-5433-ggf5, high severity, published 2026-10-11
- [Security advisory: Agent TARS CLI: unauthenticated access to the local agent server](https://www.aeon.fun/security/advisories/agent-tars-cli-unauthenticated-agent-access): GHSA-x4q8-2g66-789p, high severity, published 2026-09-23
- [Security advisory: Reasonix CLI: DNS rebinding to full agent control via serve](https://www.aeon.fun/security/advisories/reasonix-serve-dns-rebinding-agent-control): GHSA-5gf8-hg54-m9q9, high severity, published 2026-10-09
- [Security advisory: fumadocs-openapi: createProxy() open proxy and SSRF by default](https://www.aeon.fun/security/advisories/fumadocs-openapi-createproxy-open-proxy-ssrf): GHSA-7r23-3v7p-56ff, medium severity, published 2026-10-09
- [Security advisory: Maigret: stored XSS in HTML and PDF reports](https://www.aeon.fun/security/advisories/maigret-report-stored-xss): GHSA-8969-3269-9c98, high severity, published 2026-10-09
- [Security advisory: nanobot: DNS rebinding to agent control via the WebUI gateway](https://www.aeon.fun/security/advisories/nanobot-webui-gateway-dns-rebinding): GHSA-68pq-qxvf-p68w, high severity, published 2026-10-09
- [Security advisory: opencodex: memory exhaustion in the SSE decoder](https://www.aeon.fun/security/advisories/opencodex-sse-decoder-memory-exhaustion): GHSA-6f7q-px7c-fqh3, low severity, published 2026-10-09
- [Security advisory: OpenHuman: CSRF on /auth/telegram implants the attacker's session](https://www.aeon.fun/security/advisories/openhuman-telegram-auth-csrf-session-implantation): GHSA-c4m9-hw7m-vv3m, high severity, published 2026-10-09
- [Security advisory: Agent Router: MCP session hijacking via unchecked subject and default seed](https://www.aeon.fun/security/advisories/agent-router-mcp-session-hijacking): GHSA-953h-535j-jg2v, high severity, published 2026-10-07
- [Security advisory: HarnessRouter: cross-org connection hijack via a made-up bearer](https://www.aeon.fun/security/advisories/harnessrouter-console-bearer-relay-connection-hijack): GHSA-p6cq-54cg-8mpv, critical severity, published 2026-10-06
- [Security advisory: graphify: host-file read via Fortran cpp includes](https://www.aeon.fun/security/advisories/graphify-fortran-cpp-include-host-file-read): GHSA-pcc4-rvhr-2pr8, high severity, published 2026-09-27
- [Security advisory: CubeSandbox: any host path mounted into a tenant's MicroVM](https://www.aeon.fun/security/advisories/cubesandbox-host-mount-arbitrary-host-path): GHSA-8frp-jfpv-5vhv, high severity, published 2026-09-18
- [Security advisory: Capminal contracts: CAPU mint over-issuance in one call](https://www.aeon.fun/security/advisories/capminal-capu-mint-curve-over-issuance): GHSA-2qg4-mp8w-wr59, high severity, published 2026-09-06
- [Security advisory: LoopX: any website can read serve-status control-plane state](https://www.aeon.fun/security/advisories/loopx-serve-status-wildcard-cors-read): GHSA-vx2m-gpq4-8j5q, high severity, published 2026-08-12
- [Security advisory: colibri: heap overflow in the safetensors loader](https://www.aeon.fun/security/advisories/colibri-safetensors-loader-heap-overflow): GHSA-4gw4-j89j-4c8r, high severity, published 2026-08-05
- [Security advisory: Kaneo: Gitea token and webhook secret exposed to workspace members](https://www.aeon.fun/security/advisories/kaneo-gitea-integration-secrets-exposed): GHSA-43pc-xpv9-6566, high severity, published 2026-08-05
- [Security advisory: Omnigent: guardrail bypass via fail-open shell parser](https://www.aeon.fun/security/advisories/omnigent-shell-parser-fail-open-guardrail-bypass): GHSA-7mqg-cx4g-x2rf, high severity, published 2026-06-29
- [Security advisory: ds2api: WebUI path traversal and API keys in access logs](https://www.aeon.fun/security/advisories/ds2api-webui-path-traversal-api-key-logging): GHSA-rf34-c5jc-4ffw, low severity, published 2026-05-10
- [Hooks](https://www.aeon.fun/hooks): live Uniswap v4 hooks aeon generated and deployed onchain
- [Ecosystem](https://www.aeon.fun/ecosystem): agents, products, and forks built on aeon
- [More aeon products](https://www.aeon.fun/#projects): first-party tools connecting to aeon framework - OpenDia, soul.md, Minitor, univ4-hooks
- [Changelog](https://www.aeon.fun/changelog): shipped changes, synced from merged PRs
- [Transparency](https://www.aeon.fun/transparency): $aeon treasury holdings over time
- [Highlights](https://www.aeon.fun/highlights): aeon milestones as shareable stickers, one page each
- [Blog](https://www.aeon.fun/blog): essays and dispatches

## Source & packages

- [GitHub repository](https://github.com/aeonfun/aeon): the canonical source, MIT licensed
- [STRATEGY.md](https://github.com/aeonfun/aeon/blob/main/STRATEGY.md): the north-star file that steers every skill
- [OpenDia](https://github.com/aeonfun/opendia): connect your browser to AI models
- [soul.md](https://github.com/aeonfun/soul.md): the best way to build a personality for your agent
- [Minitor](https://github.com/aeonfun/minitor): monitor the current thing. Your dashboard for the internet
- [univ4-hooks](https://github.com/aeonfun/univ4-hooks): Uniswap v4 hook registry; live catalog at https://www.aeon.fun/hooks

## Company

- Maintained by Aeon Inc
- [About](https://www.aeon.fun/about) · [Contact](https://www.aeon.fun/contact) · [Privacy](https://www.aeon.fun/privacy) · [Terms](https://www.aeon.fun/terms)
- On X: https://x.com/aeonframework
- Contact: aaron@aeon.fun