Creating Your Own Bundle
Creating a bundle is straightforward - it's just a markdown file with YAML frontmatter. This guide walks you through creating your first bundle.
Quick Start: Minimal Bundle
Create a file called bundle.md:
---
bundle:
name: my-first-bundle
version: 1.0.0
description: My custom Amplifier configuration
includes:
- bundle: foundation
---
# My Custom Bundle
You are a helpful AI assistant customized for my workflow.
Run it:
amplifier run --bundle ./bundle.md "Hello!"
That's it! You've created a bundle that includes all foundation capabilities plus your custom instructions.
Step-by-Step: Building a Custom Bundle
1. Start with Foundation
Most bundles build on foundation:
---
bundle:
name: my-bundle
version: 1.0.0
includes:
- bundle: foundation
---
This gives you all core tools (filesystem, bash, web, search), standard hooks (logging, streaming UI), and the basic orchestrator.
2. Add Custom Modules
Add your own tools, hooks, or providers:
tools:
- module: tool-my-custom
source: git+https://github.com/you/amplifier-module-tool-my-custom@main
config:
api_key: ${MY_API_KEY} # Environment variable
hooks:
- module: hooks-my-notifications
source: file:///absolute/path/to/local/module
3. Override Provider
Change which AI provider to use:
providers:
- module: provider-openai
source: git+https://github.com/microsoft/amplifier-module-provider-openai@main
config:
default_model: gpt-5.1
priority: 1
4. Add Custom Agents
Define specialized agents for your domain:
agents:
- agent: my-domain-expert
file: agents/domain-expert.md
Create agents/domain-expert.md:
---
meta:
name: my-domain-expert
description: Expert in my specific domain
---
You are a domain expert specializing in...
5. Add Context Resources
Include domain knowledge and instructions:
context:
- file: context/domain-knowledge.md
- file: context/coding-standards.md
Reference them in your system prompt with @mentions:
---
# (frontmatter above)
---
# My Bundle
@my-bundle:context/domain-knowledge.md
Follow these coding standards:
@my-bundle:context/coding-standards.md
Complete Example
---
bundle:
name: data-science-bundle
version: 2.1.0
description: Data science workflow with Jupyter and plotting support
includes:
- bundle: foundation
tools:
- module: tool-jupyter
source: git+https://github.com/microsoft/amplifier-module-tool-jupyter@main
- module: tool-plotting
source: git+https://github.com/you/amplifier-module-tool-plotting@main
providers:
- module: provider-anthropic
source: git+https://github.com/microsoft/amplifier-module-provider-anthropic@main
config:
default_model: claude-sonnet-4-5
priority: 1
agents:
- agent: data-analyst
file: agents/data-analyst.md
- agent: visualization-expert
file: agents/viz-expert.md
context:
- file: context/pandas-best-practices.md
- file: context/visualization-guidelines.md
---
# Data Science Assistant
You are an expert data science assistant specializing in Python, pandas, and data visualization.
## Your Capabilities
@data-science-bundle:context/pandas-best-practices.md
@data-science-bundle:context/visualization-guidelines.md
## Your Approach
1. Understand the data first
2. Clean and validate
3. Analyze with clear explanations
4. Visualize insights effectively
Best Practices
source: URLs for modules. This makes bundles self-contained and
reproducible.
Testing Your Bundle
# Run locally
amplifier run --bundle ./bundle.md "Test prompt"
# Or add to registry
amplifier bundle add ./bundle.md
amplifier bundle use my-bundle
amplifier run "Test prompt"
Sharing Your Bundle
Once your bundle works, share it via Git:
# Push to GitHub
git init
git add bundle.md agents/ context/
git commit -m "Initial bundle"
git remote add origin https://github.com/you/my-amplifier-bundle
git push -u origin main
# Others can use it:
amplifier run --bundle git+https://github.com/you/my-amplifier-bundle@main
Advanced Topics
Examples to Study
- amplifier-bundle-recipes - Multi-step orchestration
- amplifier-bundle-lsp - Code intelligence
- amplifier-bundle-design-intelligence - Design systems
- foundation bundle.md - The foundation itself