Frequently Asked Questions

Is Agent Protocol free to use?

Yes, Agent Protocol is an open specification. You can implement it, use it, and build on it freely in both open-source and commercial projects.

Which programming languages are supported?

Agent Protocol is language-agnostic. Official SDKs exist for Python and JavaScript/TypeScript, but you can implement the REST API in any language.

Do I need to use the SDK?

No. The SDK simplifies implementation, but you can implement the REST API directly using any web framework (FastAPI, Express, Flask, etc.).

Is Agent Protocol compatible with LangChain/LlamaIndex/other frameworks?

Yes! Agent Protocol is a communication layer. You can build your agent using any framework (LangChain, LlamaIndex, custom code) and wrap it with Agent Protocol for external communication.

How do I propose changes to the protocol?

Follow the RFC process on GitHub. Community input drives the protocol’s evolution.

Can agents communicate with each other?

Agent-to-agent communication is a planned future feature. The current version focuses on human-to-agent communication.

Is there authentication/authorization built in?

The protocol doesn’t mandate a specific auth mechanism. Implementations can use API keys, OAuth, JWT, or operate without auth for local use.

What’s the difference between tasks and steps?

A task is the high-level goal you give the agent. Steps are the individual units of work the agent performs to complete that task.