Guided first install Scroll the workspace path

terminal agent-guided install 1/4

Dry-run first

Let your local agent install Spark with you.

models choose model 2/4

Spark setup

Choose how Spark thinks.

default Agent + Mission ready

telegram connect chat 3/4

Create bot with @BotFather

Paste token locally

Spark Live connected

build start building 4/4

Start chatting and building.

Install Spark. Start building.

Choose how Spark thinks, connect Telegram, then start chatting and building with your agent. Best path: let Codex, Claude Code, Cursor, or another local agent sit with you during setup.

Copy agent prompt See what it changes

1 Local agent terminal

Codex, Claude Code, Cursor, or another agent harness.

2 One model path

Subscription, API key, local model, or hosted router.

3 Telegram app

Create a bot with @BotFather; paste the token locally.

Step 1 - recommended

Let your local agent install Spark with you.

This is the easiest route today. Your local agent can choose the right installer for your computer, explain the dry-run plan, help you pick one model provider, and then connect Telegram without guesswork.

Use this with Codex, Claude Code, Cursor, or another local agent Recommended path. The agent reads the Spark docs and walks you through the install.

Install Spark from https://agent.sparkswarm.ai/install. Show me the dry-run plan first, help me choose one LLM provider, then help me connect Telegram and start chatting with Spark.

Manual fallback

Installing without an agent is possible, but more hands-on.

If you do not have Codex, Claude Code, Cursor, or another local agent harness yet, use the command for your operating system. Spark still shows the plan first and guides the setup from your terminal.

Mac / Linux / WSL Paste into Terminal

curl -fsSL https://agent.sparkswarm.ai/install.sh -o spark-install.sh
bash spark-install.sh --dry-run
bash spark-install.sh

Windows Paste into PowerShell

iwr https://agent.sparkswarm.ai/install.ps1 -OutFile spark-install.ps1
powershell -NoProfile -ExecutionPolicy Bypass -File .\spark-install.ps1 -DryRun
powershell -NoProfile -ExecutionPolicy Bypass -File .\spark-install.ps1

Tutorial

The first run should feel like this.

Keep it boring and local: read the plan, pick one model path, connect one Telegram bot, then start with a tiny message before you ask Spark to improve itself for your workflows.

Let your agent read the install page

Paste the recommended prompt into Codex, Claude Code, Cursor, or another local agent. Ask it to show the dry-run plan first.

Choose how Spark thinks

Start with one default provider for Agent and Mission. Split roles later only when you know why.

Connect Telegram safely

Create a bot with @BotFather. Paste the token only into local Spark setup; never into this website.

Start chatting and building

Send /start if Telegram asks, then a normal message, then /run say exactly OK.

Improve when ready

Once Spark is useful, ask it how it can self-improve for your specific workflows. Your agent will guide that path.

Already installed?

Update Spark.

If Spark is already on your computer, start with the normal update command. It keeps your existing setup, pulls the latest modules, and avoids making you reinstall from scratch.

Normal update spark update

Use this when Spark already works and you just want the newest fixes.

After updating spark verify --onboarding

Checks Telegram, providers, memory, Spark Live, and missions.

Installer fallback for very old Spark installs

Use this only when your local Spark does not know spark update or another command from the docs.

Mac / Linux / WSL bash spark-install.sh --upgrade-existing

Windows powershell -NoProfile -ExecutionPolicy Bypass -File .\spark-install.ps1 -UpgradeExisting

Want to preview first? Add --dry-run on Mac/Linux/WSL or -DryRun on Windows.

If anything feels stuck

Go to the exact page for the problem.

Model or provider setup Codex, Claude, OpenRouter, Z.AI, MiniMax, Hugging Face, LM Studio, Ollama. Telegram is quiet Use spark fix telegram, /diagnose, logs, and repair commands. Command reference What to run for setup, Spark Live, status, providers, security, and missions. Security and tokens What stays local, what gets checked, and how Spark avoids leaking secrets.