Quick Start Guide

Getting Started
with Amplifier

From install to productive in 5 minutes

Active

February 2026

Before You Start

What you'll need

Python 3.11+

Modern Python required. Check with python --version

API Key

Anthropic recommended (Claude). OpenAI, Azure, or Ollama also supported.

5 Minutes

That's all it takes. Seriously.

Step 1: Install

Two commands to install

# Install UV (Python package manager)
curl -LsSf https://astral.sh/uv/install.sh | sh

# Install Amplifier globally
uv tool install git+https://github.com/microsoft/amplifier

            That's it. Amplifier is now a global command. Update anytime with amplifier update.
        

Step 2: Configure

Set up your AI provider

# Interactive setup wizard (auto-runs on first use)
amplifier init

# Prompts you to:
#   Choose provider (Anthropic, OpenAI, Azure, Ollama)
#   Enter API key
#   Select default model

            Settings saved to ~/.amplifier/settings.yaml. Switch providers later with amplifier provider use.
        

Step 3: Start Chatting

Launch your first session

# Start an interactive session
amplifier run

# Or execute a single prompt directly
amplifier run "List the files in the current directory"

# Override provider/model per session
amplifier run -p anthropic -m claude-sonnet-4-20250514

            You're running an AI agent. It reads files, writes code, searches the web, and executes commands—tools are selected automatically.
        

Essential Commands

What you need to know

$ amplifier run — Start a new interactive session
$ amplifier continue — Resume your most recent session
$ amplifier resume — Pick from recent sessions to resume
$ amplifier bundle list — See available bundles
$ amplifier update — Update to the latest version

Try This

Your first productive task

# Analyze a file
> Read the README.md and summarize the key points

# Create something
> Create a Python script that calculates fibonacci numbers

# Multi-step task
> Find all TODO comments in this project,
  prioritize them, and save a report to todos.md

            Tools are automatic. File operations, web search, code execution—Amplifier selects and uses them without you asking.
        

FAQ

Common first questions

How do I switch models?

amplifier provider use to change default, or -p and -m flags per session

Where are sessions saved?

~/.amplifier/projects/<project>/sessions/ — Full history in JSONL format

How do I add capabilities?

Bundles! amplifier bundle add git+https://github.com/org/bundle

Can I use it offline?

Yes—configure Ollama as your provider for fully local models

How do I track cost?

Use amplifier run -v (verbose) to see token usage per request

Is my data safe?

Everything stays local except API calls to your chosen provider. No telemetry.

Sources

Research Methodology

Data as of: February 20, 2026

Feature status: Active — Amplifier v2026.02.18-0a861ae

Research performed:

amplifier --version → v2026.02.18-0a861ae
amplifier --help → 17 subcommands verified
amplifier run --help → flags: --bundle, --provider, --model, --mode
amplifier init --help → auto-runs on first use, supports -y for CI
amplifier bundle list → 14 bundles in registry
amplifier provider --help → subcommands: current, install, list, models, use
ls ~/.amplifier/settings.yaml → confirmed settings location
uv tool list → confirmed install method (amplifier v0.1.0)

Gaps: In-session slash commands not verified (requires interactive session). Python version requirement stated but not confirmed from pyproject.toml.

You're ready.

Explore, experiment, build.

Full Tutorial: ramparte.github.io/amplifier-tutorial
Best Practices: See the “Deliberate Development” deck
Advanced: Bundles, Agents, Recipes, Multi-Agent

Sessions are isolated—experiment freely, nothing breaks permanently

Getting Startedwith Amplifier