Your AI, handled.
Not demoed.
You already know AI can do more than chat. The problem is the last mile: setup, reliability, guardrails, and the messy reality of your email/calendar/tools. Feral Bots builds assistants that ship. Local-first when you want privacy. Always-on when you want reach.
You’ve seen the tweets. Now comes the hard part.
AI assistants are everywhere — in threads, demos, and breathless screenshots. But the thing you actually want is not a prompt. It’s a system that works on Monday morning. On your phone. With your inbox. With guardrails.
- Tool sprawl: five apps, twelve logins, zero integration.
- Fragile automation: breaks quietly, fails loudly.
- "AI" that can’t act: it talks — you still do the work.
- Privacy anxiety: you don’t want your life in someone else’s cloud.
- End‑to‑end deployment: local, cloud, or zero-cloud. We handle the stack.
- Reliability ops: keep‑alive, monitoring, upgrades, and "why is it down?" answered.
- Capability with restraint: guardrails that prevent noise, runaway spend, and accidental sends.
- A real assistant: it checks in, it remembers, it routes work to the right agent.
What you get (the parts that actually matter)
Not a chatbot. A deployed assistant with real-world capabilities we’ve already implemented in production.
24/7 assistant on Telegram / webchat
Talk to your AI where you already are. For Pro and above, we deploy an always-on Cloudflare container so it’s there at 2am, not “when your laptop is open.”
Email triage + controlled sending
Smart sorting, drafts, and safe outbound workflows. We’ve shipped programmatic outbound email via Resend with guardrails — not “copy/paste into Gmail.”
Proactive briefings
Daily/weekly briefings, upcoming-event context, and “don’t let me forget this” capture — without you babysitting it.
A team, not a monolith
Fast triage agent + deep-thinking specialist + optional researcher/coder. We run heterogeneous model fleets (Opus/Gemini/GPT-class) with sane routing.
They actually remember
Shared Exobrain memory (D1 + embeddings) means your agents can hand off work and keep context. No more “remind me again who Murat is.”
Monitoring, keep-alive, and upgrades
We’ve built keep-alive services to prevent cloud sleep, wired health checks, and shipped safe migrations. This is why it doesn’t crumble at week two.
Local-first → air-gapped
Starter can stay on your Mac. Sovereign goes further: zero cloud, on-prem hardware, optional air-gapped operation. “Trust no one” is a supported configuration.
Private networking (Tailscale + SSH)
We wire distributed agents and services into a private tailnet for secure reachability — not public endpoints with crossed fingers.
Cost + safety controls for real channels
We’ve seen what happens when bots respond too eagerly. Mention-only group behavior, approval workflows, and spend sanity checks are part of the craft.
How it works
Three steps. No ceremony. We build it, you use it, we keep it sharp.
We map your life to an assistant
Inbox, calendar, recurring workflows, and where you want the assistant to live (Mac, cloud, or sovereign). We recommend a tier and a rollout plan.
We deploy the system
Profiles, tools, skills, channels, guardrails, memory. If it’s Pro+, we deploy a Cloudflare container, wire storage, and add monitoring.
You get leverage (and we keep it alive)
Training session, playbooks, and ongoing support depending on tier. Updates and improvements don’t require you to become an MLOps person.
Pricing tiers
Clear packaging. No "contact sales." Upgrade anytime — we credit your previous setup fee.
Starter
For professionals who want a serious assistant on their Mac.
- OpenClaw profile on your Mac
- Telegram bot for mobile access
- 5–10 skills configured (email, calendar, notes, reminders, etc.)
- 2-hour hands-on training + 30-day support
- Local by default (your machine, your data)
Pro
For executives who need 24/7 access from anywhere.
- Always-on Cloudflare container deployment
- Webchat interface (custom subdomain)
- R2 storage + encrypted data at rest
- Tailscale VPN for secure access
- Monthly check-in + priority support
Full Stack
For power users who want multi-agent + shared memory.
- 2–4 agents (triage + specialist + more)
- Exobrain shared memory (D1 + embeddings)
- Agent council for complex decisions
- Custom integrations (Slack/CRM/etc.)
- On-site setup day (travel extra)
Sovereign
For when zero cloud isn’t optional.
- On-prem hardware included (2× Mac Studio class)
- Local LLM deployment (no API calls)
- Optional air-gapped mode
- You own the hardware + the system
- “Trust no one” is a supported config.
501(c)(3) organizations get 50% off setup fees. Same monthly retainer. We believe in your mission — even if we’re skeptical of the tax code.
WhatsApp, Slack/Discord, voice interface (Siri/Shortcut), inbox monitoring, browser automation, custom MCP servers, image gen, TTS. If it’s boring and repeatable, we can probably hand it to the bots.
Proof (not promises)
Some of what we’ve shipped recently — the kind of work that turns “cool” into “reliable.”
Always-on agents in Cloudflare Containers
Reproducible startup scripts + secrets, fixed boot crashes, and deployed stable container agents with real uptime expectations.
Pages deployments + custom domains
Shipped hosted dashboards with SSL-backed custom domains — so non-technical users have a control surface, not a CLI.
Cross-agent memory + council workflow
Deployed shared memory and a formal council request/response trail for high-stakes actions — transparency beats magic.
“This is the first assistant that feels like it’s on my side — not just answering prompts.”
“The system didn’t just get set up… it got maintained. That’s the difference.”
“Multi-agent + shared memory changed how fast we move. It feels unfair.”
Meet the Team
Six agents. One human. Zero excuses.
FAQ
If you’re wondering it, we’ve already heard it. (Often right before someone booked the call.)
Do I need to be technical?
No. You need to be able to describe your workflows and send a message on Telegram. We handle setup, deployment, and maintenance.
What about ongoing AI subscription costs?
Starter/Pro generally require Anthropic Max ($200/mo) or Google AI Ultra ($250/mo). We’ll tell you exactly what you need for your tier and usage. Sovereign can run with zero cloud model costs.
Is my data safe?
It depends on tier. Starter is local to your machine. Pro/Full Stack use encrypted storage in Cloudflare. Sovereign can be fully on-prem, with an air-gapped option. We never access your data without explicit permission.
Can I upgrade later?
Yes. We credit your previous setup fee toward the upgrade. Start local, go always-on later. Or go full Sovereign when it’s time.
Will the bot spam my group chat?
No. We build guardrails like mention-only response modes and approvals for sensitive actions. We’ve learned this lesson the expensive way so you don’t have to.
What’s the first step?
Book a discovery call. We’ll map your needs to a tier, answer privacy questions, and outline the build plan. If it’s a fit, we schedule the setup.