1. We provision or prep your environment
Cloud spins up automatically. On-Site and Ship-In move into scheduling and hardware handoff.
Pick your deployment, choose Mercury Credits or BYOK, connect your first channel, and get to your first real workflow fast. No fluff. Just the shortest path from checkout to a useful operator.
Cloud spins up automatically. On-Site and Ship-In move into scheduling and hardware handoff.
Use Mercury Credits for the fastest start, or connect your own provider keys if you want direct model control.
Telegram, Discord, WhatsApp, Signal, Slack, or iMessage — then give Mercury a job that matters today.
Best for founders, operators, and teams who want a private always-on agent without touching infra.
Typical time to first live session: 10–15 minutes after payment.
Best for offices or teams with stricter hardware, network, or data-handling requirements.
Best when local environment control matters more than instant deployment speed.
Best when you want Mercury running on your own box without doing the install yourself.
Best when you want ownership of the machine but not the setup headache.
If you do only these seven things, Mercury will stop feeling like AI infrastructure and start feeling useful.
Both work. The right answer is speed vs control.
Best for people who want one bill, no provider-account juggling, and the fastest setup path.
Tradeoff: less direct provider-level control.
Best for customers who already use Anthropic, OpenAI, Google, or other providers and want direct billing control.
Tradeoff: a few more setup steps up front.
Telegram or another supported channel is the only real must-have. API keys are only required if you choose BYOK.
Most customers should start simple, then add integrations after the first real win.
If you choose BYOK, Mercury never locks you to one provider. Claude is strong for polished writing and strategy, GPT is flexible, Gemini is cost-effective for lightweight checks, and Kimi often gives excellent value.
Support is included for the first 30 days. Use support@mercuryinstall.com for setup help, /support for support-hour options, or /contact if your deployment needs a human handoff.
These five steps turn your agent from "installed" into genuinely useful. Copy the prompts directly into Telegram to accelerate setup.
Start with a clean intro so the assistant understands context and tone right away.
Hey! I'm [name]. Introduce yourself and tell me what you can do for me day-to-day.
Personalization creates better answers, better drafts, and better reminders.
Let's set up my preferences. Ask me one question at a time about my role, goals, communication style, timezone, and the 3 things I want automated first. Save each answer to memory.
Prove value immediately with a real task you can verify in seconds.
Check the weather for [city] and give me a simple 3-day forecast with high/low temps and rain chance.
This is the highest-impact workflow for most users.
Set up a daily morning briefing at [time]. Include weather, today's calendar events, top important emails, and key reminders. Keep it under 200 words.
Use /compact before changing topics, "Remember that…" for permanent context, and natural-language reminders like "remind me Friday at 4 PM to send invoices."
Think in roles: your Brain handles complex conversations and writing; your Muscles handle repetitive checks cheaply and fast.
| Role | Best Fit | Typical Cost | Use Cases |
|---|---|---|---|
| Brain | Claude Opus or Kimi K2.5 | $6–$50/mo | Planning, strategy, polished writing, complex analysis |
| Muscles | Claude Haiku or Gemini Flash | $0–$3/mo | Heartbeat checks, reminders, routine summaries, lightweight monitoring |
Pro tip: Start with one solid model setup, then add routing once you see your usage patterns.
Receive one clean daily snapshot so you can act fast without opening five apps.
Send me a morning brief every day at 7:00 AM with weather, calendar, urgent emails, and my top 3 priorities.
Filter noise and only surface messages that need your attention.
Check my inbox every 2 hours, ignore promotions, flag urgent emails from key contacts, and draft replies for approval.
Track commitments and deadlines using simple natural language commands.
When I say 'add task', save it by priority and due date. Send me a 6 PM daily summary of open tasks and tomorrow's deadlines.
Get concise research summaries with links instead of reading dozens of pages.
Twice a week, monitor [industry/topic] and send me a 5-bullet summary with links and one actionable insight.
Turn rough notes into polished drafts for emails, social posts, and client updates.
Use my preferred tone (professional and concise). When I paste notes, produce a ready-to-send draft first, then a shorter version.
Treat API keys like passwords. Store in a secure password manager and never share in chats or screenshots.
Configure hard budget caps at your model providers to prevent surprise bills as automation grows.
Rotate core keys every 6-12 months and immediately after any suspected exposure.
Audit which integrations are connected and remove anything unused to reduce risk surface.
Ask your agent for a weekly "memory digest" of important preferences, workflows, and commitments so you always have a portable backup.
Launch your AI agent today and build your first high-leverage workflow in under 30 minutes.
Buy Now