Master Jarvis AI Assistant.
Command the whole system.
From your first query to advanced desktop automation, model routing, mobile control, memory, privacy modes, and real agent workflows.
Essential Commands Reference
Every powerful Jarvis interaction starts with the right command. These patterns cover 90% of daily usage.
Basic Queries
Model Selection via Prefix
Interface Modes
Privacy Overrides
Memory & Recall
Tool-Specific Queries
Pro Tip — Natural Language is Enough
Jarvis understands intent, not just keywords. You do not need to know tool names. Say "check if my API is down and open a ticket if it is" — Jarvisselects the right tools, runs them in parallel, and chains the results automatically.
Power Workflows for Real Work
Complete end-to-end workflows that show Jarvis working autonomously — not just answering questions.
Developer Daily Workflow
Morning standup to end of day
Morning Standup Prep
Ask Jarvis to summarize yesterday's commits so you can report progress without digging through git logs.
Automated Ticket Creation
Jarvis scans your codebase for TODO comments and creates properly formatted Jira tickets automatically.
Intelligent Debugging
Paste an error and Jarvis reads the relevant files, traces the stack, and suggests a targeted fix.
End-of-Day Summary
Generate a readable summary of your day, commits, and open issues for the next session.
Infrastructure & Ops Workflow
Monitor, diagnose, remediate
Natural Language Monitoring
Query container health, resource usage, and service status without memorising Docker or kubectl commands.
Network Intelligence
Check Tailscale mesh status, identify offline nodes, and get explanations of network anomalies.
Automated Diagnostics
Run a full diagnostic suite in one command — Jarvis runs tools in parallel and aggregates findings.
Runbook Execution
Jarvis remembers runbooks from past conversations. Reference them by name to execute remediation.
Research & Analysis Workflow
Research, synthesize, report
Web Research
Jarvis searches the web, scrapes relevant pages, and synthesizes findings — not just a list of links.
Cross-Session Recall
Query your own conversation history to recall decisions and research from months ago.
Multi-Model Validation
Get answers validated across multiple LLMs by routing the same question to different providers.
Report Generation
Combine filesystem data, web research, and conversation history into a structured report.
The Full Jarvis Desktop Experience
The Windows Desktop App is the recommended way to use Jarvis — no terminal needed, auto-updates, always available.
Install from Microsoft Store
One click install from the Microsoft Store. Auto-updates when new versions are released. No terminal, no setup — just install and run.
Download from Microsoft StoreStore ID: 9ND65V74RVTB
What the Desktop App Includes
Install
Download and install from the Microsoft Store. Launch Jarvis.
Add API Key
Go to Settings and add your preferred LLM API key. Groq is free to start.
Set Privacy Mode
Choose STRICT (local only), GUARDED (cloud + redaction), or RELAXED.
Start Asking
Type or speak your first query. Jarvis routes it, runs tools, responds.
Using Jarvis via Terminal (Advanced)
The CLI is ideal for power users who live in the terminal. The Desktop App and CLI share the same agent, tools, memory, and settings.
WhatsApp & Messenger Automation
Let Jarvis manage your messages intelligently — context-aware replies, scheduling, and custom rules.
WhatsApp Auto-Reply
Messenger Auto-Reply
Context-Aware
Reads message content and generates a relevant reply — not a generic canned response.
Scheduled
Set active hours. Auto-reply fires only during off-hours or when you mark yourself as busy.
Whitelist Control
Control which contacts trigger auto-reply. VIPs always get through, others get the template.
Manual Message Management
Route Every Query to the Right Model
Jarvis automatically selects the best provider — or you can override with a prefix. No config files needed.
| PREFIX | PROVIDER |
|---|---|
| claude-* | Anthropic |
| claude-code-* | Claude Code CLI |
| gpt-* | OpenAI |
| gemini* | |
| groq/* | Groq |
| ollama/* | Ollama |
| cursor* | Cursor Agent |
Routing in Practice
Set Your Default Model
Personalize Jarvis to Know You
The profile system injects your context into every conversation. Memory means Jarvis never forgets a decision.
Your Profile (~/.jarvis/profile.json)
Configure from the Desktop App or web dashboard: Settings → Profile
How Memory Works
Every conversation is saved to PostgreSQL with full-text search indexing.
Redis caches recent context for instant recall within a session.
Tool success/failure rates and error→fix mappings tracked in smart_memory.db.
Your profile is injected into every prompt with a 60-second cache — zero overhead.
Memory Commands
Configure Your Privacy Level
Three modes from fully local to cloud-optimised. Every mode is explicit — nothing surprises you.
All LLM calls go to local Ollama models only. No data leaves your device. Fernet encryption at rest. Ideal for sensitive codebases and regulated environments.
Cloud LLMs allowed but PrivacyGuard strips API keys, passwords, and PII before any send. Every external call is audit-logged. Best balance of power and privacy.
No interception layer. All providers available at full throughput. Use only for development work with non-sensitive data.
8 Patterns That 10x Your Productivity
These are not obvious from the docs. They come from using Jarvis at scale across thousands of queries.
Chain Commands Naturally
Jarvis understands "and then" instructions. One message can trigger a 5-step chain: analyze → create ticket → notify on Slack → update status → summarize.
Be Specific About Output
Tell Jarvis exactly what format you want. "In a markdown table", "as a bash script", "in 3 bullet points" — it follows formatting instructions precisely.
Reference Files Directly
Jarvis can read any file on your system. Reference them by path and it will read, analyze, and act on the content automatically.
Use Parallel Execution Explicitly
When you need multiple independent operations, say "simultaneously" — Jarvis uses asyncio.gather() and returns all results at once.
Build on Previous Responses
Jarvis retains conversation context. Follow up naturally in the REPL. "Make it shorter", "Now do the same for the API module", "Use Redis instead."
Voice Mode in the Desktop App
Enable voice mode in the Desktop App settings. Dictate queries while reading code, hear results spoken back — no keyboard interruption needed.
Project Context Auto-Loading
Register your projects once and Jarvis automatically injects the right context based on your query — no manual setup per conversation.
Ask Why It Did What It Did
If Jarvis does something unexpected, just ask. "Why did you use that tool?", "What data did you send?" — it explains its reasoning transparently.
READY TO START
You now know Jarvis.
Go build something.
Install in 3 minutes. Run your first intelligent command. Experience the difference between a chatbot and a real AI system.