A persistent AI system that runs on your machine 24/7. Powered by Claude.
Each agent has a code name and a role. You give the order. Your pack handles the rest.
Runs natively on Claude Code. Connects to Gemini, Grok, Veo, Kling, ElevenLabs, and any API you throw at it.
A self-hosted AI agent platform on your own hardware. It connects to your apps, files, browser, and terminal — and actually gets things done.
Mammals runs on your Claude subscription — not the API. Claude handles reasoning and execution. External models handle what they're best at.
Most agent platforms bill per token. Mammals is free — pay what you wish. Runs on your existing Claude Pro or Max plan.
Every other AI tool is a window you open. Mammals never closes.
Runs 24/7 on your machine. No browser tab to keep open. Starts on boot, survives restarts, picks up where it left off.
Create named specialist agents with one message. Mammals auto-routes tasks to the right agent, manages sessions, keeps them alive across restarts.
Full terminal access, file system control, browser automation via Chrome DevTools Protocol. It doesn't just suggest — it executes.
Every agent has a name, an avatar, a role, and persistent memory. Create as many as you want — each one is yours to customize.
Swipe or use arrows to browse
Via MCP (Model Context Protocol), browser automation, and direct APIs.
Mammals runs on Claude Code — it can read, write, and modify its own codebase. Build new skills, change behaviors, add capabilities. All through conversation.
"Build me a skill that checks Hacker News every morning and sends me the top 5 AI posts." Mammals writes the script, registers it, and schedules it.
Don't like how it handles something? Tell it. It modifies its own code and the change takes effect immediately.
Browse community-built skills, audit the code for safety, and install with one command. Or write your own from scratch.
you:
build me a skill that watches my front door camera and texts me if it detects motion after midnight
mammals:
[writing] scripts/door-watch.ts — camera feed + motion threshold
[registered] skill: door-alert — active midnight-6am, snapshot on trigger
Live. I'll send you a photo and alert the moment anything moves.
One command. It installs everything and opens a visual setup wizard in your browser — no terminal experience needed.
A guided web UI opens in your browser. Enter your name, connect Telegram, set up remote access — step by step. Re-access it anytime from HQ settings.
Create agents with a single message. Give each a code name and a role. They persist and remember everything.
Mammals isn't a third-party wrapper or API hack. It runs directly inside Claude Code — Anthropic's official CLI tool.
claude -p CLI, the official interfaceMammals isn't built on Claude. It's built inside Claude.
Persistent sessions, tool use, file access, shell commands — all using Claude Code exactly as it was designed. No hacks, no exploits, no workarounds.
Mammals runs on your hardware. Your data never touches a third-party server.
Everything runs on your own machine. Your files, your API keys, your conversations — none of it leaves your hardware.
API credentials are stored in a local .env file only you can access. No cloud vaults.
Mammals HQ is only accessible on your local network — never exposed to the internet. Add Tailscale for secure remote access from anywhere.
Destructive actions — deleting files, force pushes, dropping databases — always require your confirmation.
Mammals only responds to you. HQ runs on your local network, Telegram is locked to your account. No multi-tenant risk.
Every action is logged locally. See exactly what happened, when, and what the result was.
Your data is never used to train models.
Claude Code conversations fall under Anthropic's standard privacy policy. No third-party data processing.
Your pack runs on one machine. You control it from all of them.
Built-in dashboard. Chat, manage agents, monitor activity — any browser on your local network.
Encrypted mesh network between your devices. HQ becomes accessible from anywhere — no port forwarding, no exposed servers.
Message your agents from any Telegram client. Works even without VPN — always available as a backup.
Open HQ in Safari or Chrome, tap "Add to Home Screen." Full-screen, no browser chrome, instant launch — feels native on any device.
Private encrypted network between your devices. Your Mammals instance stays invisible to everyone else. Free for personal use.
You can. Here's what you're signing up for.
Session management, MCP integrations, cron scheduler, browser automation, voice, persistent memory, crash recovery, boot config — a lot of trial and error before anything works.
Session poisoning. Context window management. Prompt injection. Race conditions. Dead sessions that look alive. You don't know what you don't know — until it silently fails at 2am.
APIs change. MCP servers update. Claude models evolve. Keeping it working is its own job.
Semantic search, salience decay, episodic memory, knowledge graphs, contradiction detection — Mammals ships all of it wired together.
Skip the months of debugging and start using it today.
You can absolutely build this yourself — Mammals saves you the pain. Free to use, pay what you wish if it's useful.
Free to download. No subscriptions. No per-message fees. Support development if it's useful to you.
No subscriptions. No tiers. No gotchas.
free · suggested $20
Requires Claude Pro or Max subscription (separate, from Anthropic). Personal use license.
A persistent AI system that runs on your hardware. Your agents. Your rules. Always on.
Get Mammals — Free →macOS · Powered by Claude · Pay what you wish