- AI Engineering
- Posts
- Build Production AI Agents from One Terminal
Build Production AI Agents from One Terminal
... PLUS: Let AI Agents Pay for Their Own Tools
Sumanth P
July 09, 2026
In today’s newsletter:
Google Agents CLI: Build Production AI Agents from One Terminal
Let AI Agents Pay for Their Own Tools
Reading time: 5 minutes.
AI agents have become good at reasoning, planning, and calling APIs. But they still struggle with one practical problem. What happens when the next tool they need isn't free?
Today, that usually means creating an account, subscribing to a service, generating an API key, and adding it to the agent's runtime. The moment a payment is required, the agent stops and waits for a human.
Apify's integration with the x402 protocol, built on x402 by Coinbase, removes that bottleneck. AI agents can now discover, pay for, and run thousands of Apify Actors without an Apify account, subscription, API key, or manual approval.
The Missing Piece
HTTP has had a reserved status code for 402 Payment Required since its earliest specifications, but it was never formally adopted.
The x402 protocol gives that status code a practical purpose.
When an agent calls a paid endpoint, the server responds with an HTTP 402 containing the payment details. The agent signs a payment from its wallet using USDC on Base, retries the request, and the tool executes automatically. The payment becomes part of the API call instead of a separate checkout flow.
Why This Integration Matters
Until now, x402 mostly connected agents to individual paid endpoints.
Apify expands that ecosystem by bringing ~20,000 eligible web automation tools into the protocol through its marketplace of Actors. These serverless programs cover tasks such as web scraping, browser automation, and data extraction.
That means an autonomous agent can discover a suitable Actor, pay for a single execution, retrieve the results, and continue its workflow without any service-specific credentials.
Key Capabilities
Discover and run thousands of Apify Actors autonomously.
Pay per execution using USDC on Base through the x402 protocol.
No Apify account, subscription, or API key required.
Access supported Actors through the Apify MCP server or directly over HTTP.
Use Coinbase's Agentic Wallet CLI (
npx awal) or Apify'smcpcclient to handle wallet management and payment automatically.
What This Enables
Imagine an agent researching competitors.
It identifies that it needs pricing information, customer reviews, recent social media posts, and Google Maps business data.
Instead of relying only on the tools it was configured with beforehand, the agent can discover the appropriate Apify Actors, pay for each one when needed, collect the data, and move on to the next task.
The workflow stays autonomous from start to finish.
Building an AI agent doesn't end when the code runs.
You still need to evaluate it, deploy it, monitor it, and make it available to other users. Most developers switch between multiple tools to get through those steps: an editor for development, a terminal for scaffolding, evaluation scripts, cloud services for deployment, and another interface to publish the finished agent.
Google's Agents CLI brings those stages into a single workflow. More importantly, it teaches your coding agent how to perform them for you.
More Than a CLI
At its core, Agents CLI is a command-line tool for building agents with Google's Agent Development Kit (ADK). It can scaffold projects, run evaluations, deploy agents, and publish them to Gemini Enterprise.
The interesting part is how it works with AI coding agents.
Running agents-cli setup installs a collection of ADK-specific skills into tools like Claude Code, Codex, Gemini CLI, and other compatible coding assistants. Instead of copying commands from documentation, you can ask your coding agent to scaffold a project, generate evaluation cases, deploy an agent, or configure observability using natural language.
The Skills It Learns
The installed skills cover the entire development lifecycle.
Your coding agent learns how to:
Scaffold new ADK projects or upgrade existing ones.
Build agents using ADK patterns for tools, orchestration, callbacks, and state management.
Generate and run evaluation datasets, including LLM-as-a-judge scoring.
Deploy agents to Google Cloud.
Register deployed agents with Gemini Enterprise.
Configure observability with Cloud Trace, logging, and third-party integrations.
Rather than memorizing dozens of commands, the coding agent has the documentation and workflows available as reusable skills.
One Workflow
A typical development cycle looks like this:
Install Agents CLI and the ADK skills.
Ask your coding agent to scaffold an agent from a template.
Build and iterate on the implementation.
Run evaluations and review the scorecard.
Deploy the agent to Google Cloud.
Publish it to Gemini Enterprise so others in your organization can discover and use it.
Each step is available through the CLI, but the coding agent can orchestrate the workflow for you.
The Takeaway
AI coding agents are becoming more capable, but they still depend on the knowledge they're given.
Agents CLI packages Google's ADK documentation, workflows, and best practices into reusable skills that your coding agent can load on demand. Instead of repeatedly explaining how to scaffold, evaluate, or deploy an agent, those instructions become part of the agent's working knowledge.
It's a good example of where AI development is heading: not just better models, but better ways to teach models how to complete specialized engineering workflows.
That’s all for today. Thank you for reading today’s edition. See you in the next issue with more AI Engineering insights.
PS: We curate this AI Engineering content for free, and your support means everything. If you find value in what you read, consider sharing it with a friend or two.
Your feedback is valuable: If there’s a topic you’re stuck on or curious about, reply to this email. We’re building this for you, and your feedback helps shape what we send.
WORK WITH US
Looking to promote your company, product, or service to 200K+ AI developers? Get in touch today by replying to this email.