Learn the Open Agent Format through progressively complex examples with downloadable sample agents.
The Open Agent Format (OAF) is an open standard for packaging AI agents, making them portable across different agent harnesses and platforms.
These samples demonstrate the different composition patterns available in OAF. Each builds on the previous, introducing new capabilities.
Download any sample as a .zip file, extract it, and explore the structure. All samples are OAF-compliant and ready to import into any compatible harness.
The simplest possible agent—just an AGENTS.md file. No skills, no MCPs, no sub-agents. Perfect for single-purpose conversational agents.
Generates personalized motivational affirmations based on user goals and current mindset. A focused agent that needs nothing beyond its core instructions.
daily-affirmation/
└── AGENTS.md
Adds a skills/ directory containing reusable knowledge modules. This example includes scripts/ for structured output formatting.
Transforms meeting notes or transcripts into structured summaries with action items, decisions, and next steps. Uses a Python script to ensure consistent output formatting.
scripts/ pattern—executable code that the agent can invoke for precise formatting or calculations.
meeting-summarizer/
├── AGENTS.md
└── skills/
└── summarization/
├── SKILL.md
├── resources/
│ └── output-format.md
└── scripts/
└── formatter.py
Introduces mcp-configs/ for connecting to external services via the Model Context Protocol. Combines domain knowledge with API integrations.
Finds recipes based on available ingredients while respecting dietary restrictions. Uses a skill for dietary knowledge and an MCP server to query a recipe database.
recipe-finder/
├── AGENTS.md
├── skills/
│ └── dietary-preferences/
│ ├── SKILL.md
│ └── resources/
│ └── substitutions.md
└── mcp-configs/
└── recipe-api/
├── ActiveMCP.json
└── config.yaml
Multiple independent top-level agents bundled in a single package. Each agent has its own AGENTS.md in a separate directory. Ideal for distributing related agents as a toolkit.
A toolkit of specialized travel agents that work together: a coordinator, flight researcher, hotel researcher, activity curator, and logistics planner. Each is an independent agent with its own skills.
travel-research/
├── trip-coordinator/
│ ├── AGENTS.md
│ └── skills/travel-planning/
├── flight-researcher/
│ └── AGENTS.md
├── hotel-researcher/
│ └── AGENTS.md
├── activity-curator/
│ ├── AGENTS.md
│ └── skills/destination-japan/
└── logistics-planner/
├── AGENTS.md
└── skills/budget-optimization/