Introducing A2A Cloud Academy: four free courses from a2a init to production

Introducing A2A Cloud Academy: four free courses that take you from a2a init to production

Today we're launching A2A Cloud Academy — a free learning hub at learnhouse.a2acloud.io with four in-depth courses covering the entire a2a cloud journey: your first deployed agent, the skill runtime, production operations, and connecting agents to everything else through frontends, OpenAPI, and MCP.

Documentation tells you what every command does. The Academy teaches you how to *think* in agents: why skills are typed contracts, why authority travels as signed grants instead of shared API keys, and why every run ends with a receipt you can audit. Fifty-three lessons, all grounded in the same platform you deploy to — every command, flag, and code sample comes straight from the shipping product.

The four courses

1. Your First Agent: Zero to Deployed

The beginner path. Install the CLI, scaffold a Python agent, write your first @skill, test it in the local dev console, and ship it with one command to a live *.a2acloud.io subdomain — A2A protocol, MCP endpoint, TLS, and docs included.

a2a init hello-agent
cd hello-agent
a2a dev        # local loop at http://127.0.0.1:8000
a2a deploy     # live at https://hello-agent.a2acloud.io

Start here if you've never deployed an agent before. By the last lesson you'll have a live agent and your first Workspace chat with it.

Take the course →

2. Skills & the Runtime Context: Building Agents That Do Real Work

The deep dive. Everything the ctx runtime hands your skill: streaming progress, pausing for human answers with ctx.ask, negotiating file access mid-run with ensure_read and ensure_write, BYOK LLM calls through ctx.llm, the microVM sandbox, workspace file exchange, memory, and calling other agents with scoped grants.

• How Ed25519-signed grants carry authority — audience-bound, time-limited, auditable
• Why scope negotiation beats asking for everything up front
• How every run produces a signed receipt: caller, tool calls, cost, outcome

Take the course →

3. Running Agents in Production: Keys, Schedules & Observability

The operations manual. Bring your own LLM keys (OpenAI, Anthropic, Gemini, Groq, Mistral, and more via the LiteLLM catalog), store agent credentials in the secrets manager, put agents on cron with Schedules, and read the Activity timeline, receipts, and proofs when something needs explaining. Plus SSH dev boxes — a2a ssh my-agent drops you into a throwaway shell with the repo pre-cloned, straight from VS Code or Cursor.

Take the course →

4. Frontends, OpenAPI & MCP: Connecting Agents to Everything

The integration playbook. Ship a React UI packed with your agent, turn any OpenAPI spec into a deployed agent with a2a openapi generate, export your skills as an OpenAPI 3.1 contract or a typed TypeScript client, wire your agent into Claude as MCP tools with a2a mcp-url, import agents you already run elsewhere, and compose meta-agents that orchestrate sub-agents with budgets and grants.

Take the course →

How the courses are built

Every lesson follows the same shape: the concept and *why it exists*, real code you can paste, the pitfalls we see people hit, and a short "check yourself" before you move on. Lessons run 300–700 words — long enough to explain, short enough to finish on a coffee break.

The source of truth is the platform documentation at docs.a2acloud.io, so what you learn is what ships. When the platform moves, the Academy moves with it.

Start learning

• Browse all courses: learnhouse.a2acloud.io
• New to the platform? Begin with Your First Agent: Zero to Deployed
• Prefer raw reference? The docs remain at docs.a2acloud.io, including llms.txt endpoints for your AI tools

Everything is free. Deploy your first agent today — the free tier includes a hosted agent, 10,000 invocations a month, and a signed receipt for every run.