5dive

spawn and manage AI agents that talk to each other.

MIT. The same binary that runs every agent on 5dive.com — no open-core split. Skip the ops with the managed VM, or run your own.

Quickstart

# 1. install
curl -fsSL https://install.5dive.com | sudo bash

# 2. create your first agent (optional interactive wizard)
5dive init

Talk to an agent from your phone — wire it to a Telegram bot (BotFather gives you the token):

5dive agent create my-agent --type=claude --channels=telegram --telegram-token=<token>
5dive agent pair   my-agent --code=<pairing-code>

For scripted / CI setup, see 5dive init --help.

How it works

Each agent is its own Linux user running an official agentic AI CLI (claude, codex, gemini, ...) as a systemd service. They reach each other by invoking the same 5dive CLI — that is the bus. Channels like Telegram attach per agent.

            one host
 ┌──────────────────────────────────┐
 │  coder      writer       pm      │
 │ (claude)   (gemini)    (codex)   │
 │    │          │           │      │
 │    └────  5dive CLI  ─────┘      │
 │       send · ask · logs          │
 └──────────────────────────────────┘
        ↕ Telegram / Discord
        (attach per agent)

No broker, no protocol, no orchestrator. Shared filesystem, shared CLI.

Why 5dive

A team that works without you. Multiple agents on one host, coordinating with each other.

Runs as a service, not a session. Your agents stay alive when you close the terminal. Message them from Telegram any time.

Every major agentic AI CLI. claude, codex, gemini, hermes, openclaw, opencode — under one team.

A subscription that's yours. Official claude CLI on your own Pro/Max — no middleman, no OAuth proxy, Anthropic-policy safe.

Safe by default. Each agent is its own Linux user under one of three isolation tiers — sandbox an agent and it can't read your home dir or sudo your box.

Want a dashboard?

The CLI is the OSS surface — every verb here, every agent, every host, all driven from /usr/local/bin/5dive.

If you'd rather click than ssh, 5dive.com is the managed version: same CLI under the hood, but the VM, hardening, backups, and dashboard are run for you.

Agent types

hermes and openclaw are community-built harnesses that can route to many providers (OpenRouter, Anthropic, Google, Moonshot, etc.). As of April 4, 2026, Anthropic and Google no longer permit routing consumer subscription OAuth (Claude Pro/Max, Gemini) through third-party harnesses — for that work, use the official claude or gemini types with your own API key. Background: We Ditched OpenClaw for Claude →.

Commands at a glance

5dive agent list / create / start / stop / restart / rm
5dive agent send <name> <text>
5dive agent ask  <name> <text> [--timeout=120]
5dive agent logs <name> [--follow]
5dive agent <name> tui

5dive account   add / login / list / show / rename / remove
5dive auth      set / login / status     # lower-level; account is the human path
5dive skill     add / list / remove
5dive doctor [--repair] [--json]
5dive watch                              # htop-style live view
5dive compose up / down / ps             # declarative agents via 5dive.yaml

Full flag reference: 5dive --help (or 5dive <verb> --help). Machine-readable output on any command via --json.

Accounts (shared auth profiles)

One sign-in, many agents:

5dive account add   work
5dive account login work --type=claude
5dive agent create agent-a --type=claude --auth-profile=work
5dive agent create agent-b --type=claude --auth-profile=work

Rename or rotate the account, every bound agent rebinds automatically.

Isolation tiers

5dive agent create my-agent --type=claude --isolation=sandboxed

No middlemen

5dive runs on your server. Auth tokens go to model providers directly — never to us. No telemetry, no error reporting, no usage data leaves the box. Each agent is one Linux user with its own login.

Long form: your auth tokens don't touch us →.

Securing your server

5dive runs agents with shell access. Standard hygiene applies:

• patch the OS (unattended-upgrades)
• SSH key-only, no root login
• firewall default-deny
• per-agent isolation tiers
• Telegram bot allowlists

Baselines: devsec.os_hardening · Lynis · fail2ban. Or skip the checklist — 5dive.com handles it.

Other paths

Docker — kick the tires without a host install:

docker build -f docker/Dockerfile -t 5dive .
docker run -d --name 5dive-demo --privileged 5dive
docker exec -it 5dive-demo bash

Offline / air-gapped — install.sh reads from $REPO (default GitHub raw). Override with REPO=file:///path/to/local/tree and pre-install apt deps. The fetched files are listed at the top of install.sh.

Context rot — long sessions degrade. Restart daily via cron:

0 4 * * * curl -fsSL https://install.5dive.com | bash -s -- --upgrade && systemctl restart '5dive-agent@*.service'

Claude-runtime agents keep project memory under ~/.claude/projects/<dir>/memory/ across restarts — session resets, knowledge stays.

For your AI agent

If you already use Claude Code / Codex / Gemini / opencode, paste this prompt — your agent installs 5dive, learns the skill, then keeps managing agents through chat:

Install 5dive on this Linux host so I can use you to manage 5dive agents.

1. Run the installer (idempotent — safe to rerun):
   curl -fsSL https://install.5dive.com | sudo bash
2. Confirm: `5dive --version` prints "5dive 0.1.x".
3. Install the 5dive-cli skill — replace <runtime> with one of
   claude-code, codex, gemini-cli, hermes-agent, openclaw, opencode:
   npx -y skills add https://github.com/5dive-com/skills --skill 5dive-cli --agent <runtime> --yes
4. Tell me to restart so the skill loads, then ask which agent to create first.

Installing onto a remote VM over SSH? Same prompt, prefix the install line with ssh -t <user@host>. Install the skill on the laptop where you're issuing ssh from, not the remote. Use ssh -t for anything needing a TTY (e.g. 5dive agent auth login).

JSON output

Every command accepts --json. Output is {ok:true,data:...} on success or {ok:false,error:{code,class,message}} on failure. Exit code matches error.code so shell pipelines branch without parsing. Progress lines stay on stderr; stdout is always valid JSON.

{ "ok": true,  "data": [ {"name": "main", "type": "claude", "active": "active"} ] }
{ "ok": false, "error": { "code": 4, "class": "not_found", "message": "no agent named 'foo'" } }

Requirements

• Linux with systemd (Ubuntu 22.04+ recommended)
• root for install (installer apt-installs jq, tmux, and other deps)

No systemd / no root / not Linux? Use the Docker image.

Contributing

See CONTRIBUTING.md. The 5dive bundle at the repo root is built from src/ via ./build.sh; CI enforces no drift.

License

MIT — see LICENSE.