Architecture · Advanced
Agent Protocols
Markdown for Machines
How AI agents communicate and understand context. An overview of AGENTS.md, Developer Knowledge API, A2A, MCP, and AP2 standards.
01
The 'Markdown for Agents' Era
While there isn't a single universal 'Markdown for Agents' protocol, major players like Google are backing a suite of standards that use Markdown to optimize AI agent performance, context retrieval, and interoperability.
“We are moving from human-readable documentation to machine-readable briefing packets.”
02
AGENTS.md (Coding Context Standard)
Unlike README.md (which is for humans), AGENTS.md is an open-source standard designed specifically to provide context to AI coding agents. It acts as a machine-readable briefing packet containing build steps, testing workflows, and project-specific conventions.
- Usage: Create an AGENTS.md or GEMINI.md file at your project root to guide the agent's behavior and style.
- Ecosystem: Supported by major tools including Google's Gemini Code Assist, Cursor, GitHub Copilot, and Claude Code.
03
Developer Knowledge API
Google launched the Developer Knowledge API to serve official documentation (like Android, Firebase, and Google Cloud) directly as Markdown instead of HTML.
01
Source of Truth
Allows AI agents to retrieve the programmatic source of truth.
02
Token Efficiency
Reduces token waste by up to 80% compared to parsing standard web pages.
03
Accuracy
Ensures the agent always has the most current framework information.
04
Agent-to-Agent (A2A) Protocol
Developed by Google and donated to the Linux Foundation, the A2A Protocol uses Markdown for 'user experience negotiation' between different AI agents.
- Capability: Allows two different agents to agree on a communication format (e.g., requesting data in Markdown for easier integration).
- Collaboration: Facilitates 'Agent Cards' where agents list their skills and endpoints to discover and interact with each other autonomously.
05
Related Connectivity Protocols
The ecosystem is rapidly expanding with specialized protocols for tool use and transactions.
| Protocol | Purpose | Key Capability |
|---|---|---|
| Model Context Protocol (MCP) | External Connectivity | Open standard for connecting agents to external tools and data sources. |
| Agent Payments Protocol (AP2) | Autonomous Transactions | Secure standard for agents to initiate transactions (e.g., booking flights or checkout). |