Profile
Back to NewsBack
GitHub Trending 13 min
Reader Mode
chenhg5/myco: Connect Claude Code, Codex and any agent CLI across machines. Agents find each other, delegate tasks and reply through one hub. Single Go binary, one-line setup.

chenhg5/myco: Connect Claude Code, Codex and any agent CLI across machines. Agents find each other, delegate tasks and reply through one hub. Single Go binary, one-line setup.

19 hours ago

myco

All your AI agents, one app.

Use the Claude Code, Codex and other agents you already pay for — start a conversation, run several at once,
reach your other computers and your phone, put them on a schedule — all from one app.

Release</a> Build</a> !Platforms LINUX DO</a>

Download for Mac · Windows (preview) · Linux · All downloads

English · 中文 · Install Guide · Video tutorial · Apps · Architecture · HTTP API

https://github.com/user-attachments/assets/76d235fc-1c4a-4f7c-88c1-ee058f4db1da

The myco app: a project is your chat with its coordinator; its members and computers are beside it

What myco solves

Your agents are good on their own; the trouble is at the edges. So you end up as the message bus, copying prompts and pasting results between them. myco connects them into one network:

  1. A cloud agent uses your computer. A claude.ai session or a VPS agent runs a command on your machine, or hands the task to the Claude Code or Codex already set up there, and gets the result back in the same conversation. No open ports; only the agents you allow; every command recorded. How ↓
  2. Your agent asks a teammate's agent. It sends the task with its context straight to theirs and gets the answer back. How ↓
  3. Your own agents work as a team. Claude Code on your laptop, Codex on a server, a cloud session: they find each other, delegate, and you watch it all in one app. How ↓

Fastest: let your agent set it up

Already using Claude Code, Codex or another AI that runs commands? Send it this sentence. It reads the setup guide, starts a hub, connects the app, adds your computers and agents, and sorts out problems on its own, asking you whenever something is yours to decide:

Set up myco for me: read https://raw.githubusercontent.com/chenhg5/myco/main/docs/agent-setup.md (if it does not load, https://cdn.jsdelivr.net/gh/chenhg5/myco@main/docs/agent-setup.md) and follow it. Ask me before deciding anything for me, before using sudo or changing system settings, and tell me the result when you are done.

The app's first screen has the same sentence (Let my agent set it up).

Download

| Platform | Download | |---|---| | Mac | myco.dmg: Apple silicon and Intel, signed and notarized. Updates itself. | | Windows (preview) | myco-app-windows-amd64.exe (ARM). Not code-signed yet: on first run choose More info → Run anyway. | | Linux | myco-app-linux-amd64 (ARM): opens the app in your browser (X11, Wayland, WSL). Details | | iPhone (beta) | Join the TestFlight beta (iOS 17 or later). | | Servers | The command-line tool — see Prefer the command line? |

Get started in three steps

1. Open the app and get a hub, the small server your agents connect to: Start on this computer with one click, or Connect an existing Hub with its app link. To run one on your server instead (it prints the app link at the end):

curl -fsSLO https://raw.githubusercontent.com/chenhg5/myco/main/deploy/install-hub.sh && sudo bash install-hub.sh --domain auto

2. Add your agents. Click Add an agent: pick a Claude Code or Codex session found on this computer, start a new one on any connected computer, or paste the sentence the app gives you into an agent on another computer and it joins by itself.

Add an agent: the Claude Code and Codex sessions found on this computer

3. Put them to work. Chat with an agent, or have one involve another: “ask the build server's Codex why CI fails”, “run the tests on my Mac”. You see each step as it happens and can correct or stop it.

A project chat with a computer's terminal open beside it, a command's output coming in

Apps

We are rebuilding the apps we use every day on Mac, Windows and phone so that agents can use them too: you and your agent work on the same page, the agent fills it in or edits it, and you see it change live. They open in the myco app under Apps, and new ones are added release by release. Some of them:

Motion video
Motion video
Promo and short videos: describe it and the agent writes each scene; add music and voice-over, export MP4.
Design studio
Design studio
Web pages, app screens, posters and slides on an editable canvas, iterated with the agent.
Meeting notes
Meeting notes
Record a Zoom or Feishu call, transcribe it on your computer, and let the agent write the minutes.
Disk cleanup
Disk cleanup
See what fills your disk, ring by ring, and move what you don't need to the Trash.
Remote desktop
Remote desktop
See and control your Mac or Windows PC from another computer or your iPhone.
Screen recorder
Screen recorder
Record the screen, then trim, split, add text and zoom, and export MP4 locally.
Stock watch
Stock watch
A shared watchlist for China, Hong Kong and US stocks with candlestick charts and notes.
Photo editor
Photo editor
Crop, adjust, filters, text and stickers; the agent can make the edits you describe.
Issue tracker
Issue tracker
Track issues Linear-style, assign one to an agent, and watch it update the status and comment.

See all 26 apps → (docs, notes, sheets, slides, a code editor, a whiteboard, video editing, games and more)

Three ways to use it

1. Let a cloud agent use your computer

The agent runs in the cloud; the repo, tools and logins are on your computer. From the same conversation it can run a command on a shell node, or hand the task to the Claude Code or Codex already set up on your computer:

myco exec --to my-mac-shell --cmd "cd ~/code/app && git status && npm test"
myco message send --to my-mac-claude --wait 15m --body "Run the migration against the local DB and tell me what changed"

Add a shell node in the app (Add a computer), choose a folder and which agents may use it. The boundaries:

  • Your permissions, not a sandbox: commands run as your user, so allow only agents you trust.
  • Only the agents you name, checked on every command; optionally only listed commands.
  • Outbound only: your computer opens no ports. Every command and its output are recorded.
  • Works on Mac, Linux (sh) and Windows (PowerShell). Tested with claude.ai Claude Code sessions and agents on servers. Details: INSTALL.md §4.10.

2. Hand work to a teammate's agent

Share one hub with your team (Settings → Members → Invite a member). Then: “ask Matthew's backend agent which fields the orders API returns”, and your agent gets the answer from theirs.

  • Owners decide who may send work to their agents.
  • The receiving agent knows who is asking, and that it is a teammate, not its owner.
  • End-to-end encryption is available for team hubs: not even the hub can read messages.

3. All your agents in one network

Claude Code, Codex, claude.ai sessions, Pi, Kiro, Gemini CLI and more (supported agents):

  • They find each other and delegate, one to one or to a whole group, and get the results back.
  • They keep their memory: every task resumes the agent's own session.
  • You see and steer it all in one app: who is working, every conversation and exchange, on all your hubs.
A member's card
Every agent has a card: what it does, where it runs, what it did lately.
Members
Members and computers in the workspace.

Myco or Raft?

Raft (formerly Slock) is a team workspace where people and agents talk in channels and take tasks together. Choose Raft for that. Choose myco if you already have agents where you work (Claude Code and Codex sessions, a claude.ai session, a teammate's agents) and want to connect those: they join as they are, work happens from their own conversations, and you run the hub yourself.

Prefer the command line?

Everything the app does is also a myco command, handy on servers and for agents.

curl -fsSL https://raw.githubusercontent.com/chenhg5/myco/main/deploy/install-client.sh | bash

| Start a hub | Command | |---|---| | On your VPS | sudo bash install-hub.sh --domain auto (real HTTPS via sslip.io, no domain needed) | | On your VPS, behind the nginx / 宝塔 / 1Panel already on port 443 | sudo bash install-hub.sh --behind nginx --domain hub.example.com (--behind panel prints the settings to fill in) | | On your VPS, no open ports (Cloudflare Tunnel or Tailscale Funnel) | sudo bash install-hub.sh --tunnel cloudflare --hostname hub.example.com (or --tunnel tailscale) | | On your computer, via Cloudflare Tunnel | myco hub up --tunnel cloudflare --hostname hub.example.com | | On your computer, via Tailscale Funnel | myco hub up --tunnel tailscale |

Add a machine with curl -fsSL https:///j/ | bash, add a session by telling it join myco: https:///j/, then:

myco agent list --online
myco message send --to build-backend --body "Fix the failing tests in ./pkg/api" --wait 15m
myco broadcast send --selector "group=review" --body "Review PR #42" --wait 30m
myco exec --to build-shell --cmd "df -h"
myco upgrade

More setups (claude.ai cloud sessions, exec nodes, aliases, webhooks, GitHub wake-ups) are in INSTALL.md (Chinese; your agent reads it fine).

All features

  • One line per machine. curl … | bash installs myco, enrolls the machine and starts a background runner. No tokens to copy.
  • No VPS required. Run the hub on a VPS, or on a computer you already have behind Cloudflare Tunnel or Tailscale Funnel.
  • Spawn agents remotely. Start a new Claude Code or Codex session in a folder on another machine. That machine's own rules decide who may, where, and with what permissions.
  • Bring any CLI. Built-in support for Claude Code, Codex, Pi, Kiro, Cursor, Gemini CLI, OpenCode, Qoder, Qwen Code, Kimi, GitHub Copilot and more, plus a template for anything else (full list).
  • Delegation that scales. 1:1 tasks with --wait, or broadcast by label (group=review,machine=vps) and track every reply.
  • Exec nodes. Run a command on a remote machine and get its output, like ssh. Senders are allow-listed and every command is audited.
  • Automations. The hub sends an agent a task on a schedule (cron, in your time zone) or when a webhook is called, e.g. a daily report or a CI failure, while your computers are off. myco automation create, or the app.
  • Attachments. Send files with --attach. Only the recipient can download them, and each one is checked with SHA-256.
  • Safe by default. Every agent has an allow-list of who may send it work; messages are leased, redelivered and idempotent. myco won't interrupt a session you are typing in.
  • A CLI agents can drive. JSON output, structured errors with a suggestion field, distinct exit codes, --dry-run, and myco schema for introspection.
  • One static Go binary. The hub, CLI and runner are the same myco program. SQLite storage, no cgo.

How it works

flowchart TB
  subgraph HUB["myco hub"]
    direction LR
    REG["Registry<br/>names, labels, aliases, allow-lists"]
    BOX["Inboxes and threads<br/>leases, retries, attachments"]
    WAKE["Wakers<br/>webhook, Claude routine, GitHub"]
  end

subgraph MAC["Your Mac"] direction TB R1["myco runner"] -.-> S1["Claude Code session"] R1 -.-> S2["Codex session"] end

CLOUD["claude.ai cloud session<br/>myco inbox wait"]

subgraph VPS["A VPS or WSL box"] direction TB R2["myco runner"] -.-> S3["Codex session"] R2 -.-> S4["exec node: shell"] end

UI["myco app<br/>Mac / Windows / Linux"] EXT["Agents woken by push<br/>webhook, routine, GitHub Actions"]

BOX <-->|"long-poll / reply"| R1 BOX <-->|"long-poll / reply"| CLOUD BOX <-->|"long-poll / reply"| R2 UI <-->|"console token"| REG WAKE -->|"push wake-up"| EXT

  • Hub: the only server (a VPS, or your computer behind a tunnel), one SQLite file.
  • Runner: one per machine. When a message arrives it resumes that agent's own session and sends the answer back.
  • Cloud sessions wait with myco inbox wait; others can be woken by a webhook or a GitHub event.
  • Everything connects out to the hub over HTTPS: no public IP, no open ports. More in docs/architecture.md.

Supported agents

Every agent below keeps its own session: each message resumes the same conversation, so context carries over. List what's installed on a machine with myco harness list.

| Agent | Command myco runs | --harness | Support | |---|---|---|---| | Claude Code | claude | claude | Native: stable | | Codex CLI | codex | codex | Native: stable | | Pi | pi | pi | Preset: stable | | Kiro | kiro-acp | kiro | Preset: stable | | Cursor CLI | cursor-agent acp | cursor | ACP: experimental | | Gemini CLI | gemini --acp | gemini | ACP: experimental | | OpenCode | opencode acp | opencode | ACP: experimental | | Qoder CLI | qodercli --acp | qoder | ACP: experimental | | Qwen Code | qwen --acp | qwen | ACP: experimental | | Kimi Code | kimi acp | kimi | ACP: experimental | | goose | goose acp | goose | ACP: experimental | | GitHub Copilot | copilot --acp | copilot | ACP: experimental | | Auggie | auggie --acp | auggie | ACP: experimental | | CodeBuddy Code | codebuddy --acp | codebuddy | ACP: experimental | | Kilo Code | kilo acp | kilo | ACP: experimental | | Cline | cline --acp | cline | ACP: experimental | | Grok | grok agent stdio | grok | ACP: experimental |

Native has a dedicated adapter; Preset is a built-in command template; ACP uses the Agent Client Protocol, with every step shown live. Experimental ones have not had long real-world use yet; reports welcome. Kiro needs uv tool install kiro-acp-gateway.

Other ways to connect:

| | How | Use it for | |---|---|---| | Any ACP agent | --harness acp with runner.binary and runner.args (e.g. opencode + [acp]) | CLIs that speak ACP, with the live view | | Any other CLI | --harness template: declare the command line and output format in the config | CLIs without a preset, no code needed | | Any command | --harness shell: the prompt arrives on stdin, stdout is the reply | Scripts and small tools | | Exec node | --harness exec: runs each task as a shell command (allow-listed senders and commands) | Remote commands, like ssh | | claude.ai cloud sessions | No runner: the session waits with myco inbox wait | Cloud agents you can't start from a terminal | | Push-woken agents | webhook, claude_routine, github_dispatch, github_issue | Agents that live elsewhere, e.g. Codex replying on GitHub. See docs/delivery.md |

Docs

| Doc | What it covers | |---|---| | INSTALL.md | Hub deployment, join links and per-agent setup | | docs/apps.md | Every app, with screenshots | | docs/architecture.md | Concepts, message state machine, runner behavior and code layout | | docs/delivery.md | Wake-up modes and enrollment for cloud sessions | | docs/api.md | HTTP API | | skill/myco/SKILL.md | Skill file that teaches agents how to use myco | | CONTRIBUTING.md | Building myco and contributing |

Community

Questions, ideas, or curious how others use myco? Scan to join the WeChat group (the code is renewed every few days; this file always has the latest):

myco WeChat group

Bugs and feature requests: GitHub Issues.

Follow the author on X: @chg80333. For business, add mongorz on WeChat, saying who you are and what you have in mind.

Why "myco"

myco- is Greek for fungus. Under a forest, fungal threads link trees into a "wood wide web" that carries nutrients and warnings between them. myco does the same for agents: it doesn't think for anyone, it only connects and delivers.

License

MIT

If myco saves you from copy-pasting between agents, a star helps others find it.

Chat with me