Guide

How to use SARTHI

From install to your first conversation in under five minutes — then the commands you'll use every day.

Step 1

Install

One line on Linux, macOS, WSL2, or Termux:

curl -fsSL https://sarthi-agent.vercel.app/install.sh | bash

On native Windows, run this in PowerShell:

iex (irm https://sarthi-agent.vercel.app/install.ps1)

The installer handles everything — Python, Node.js, and all dependencies. No admin rights needed. Prefer source? Clone the sarthi-agent folder and run uv pip install -e ".[all]".

Step 2

Set up your model

sarthi setup    # full setup wizard
sarthi model    # or just pick a provider & model

SARTHI works with any LLM provider — see the LLM API configuration guide for OpenRouter, Nous Portal, OpenAI, and more.

Step 3

Start chatting

sarthi          # interactive terminal UI
sarthi gateway  # connect Telegram, Discord, Slack, WhatsApp…
sarthi doctor   # diagnose any issue

Inside a conversation, slash commands give you control — they work the same in the terminal and on messaging platforms:

everyday commands

/new         start a fresh conversation
/model       switch provider or model
/skills      browse your agent's learned skills
/personality set a personality
/compress    compress context, keep going
/usage       see token usage
/undo /retry fix the last turn

Going further

Let it work while you're away

Ask SARTHI to schedule anything in plain language — it has a built-in cron system with delivery to any connected platform:

❯ "every morning at 7, summarize my inbox and send it to telegram"
✓ Scheduled: morning-inbox · daily 07:00 → Telegram

It can also spawn parallel subagents for big jobs, search its own past sessions, and write scripts that call its tools directly. The more you use it, the more it learns — that's the point.