Claude Agent SDK Credit Pool: A Plain Guide to the June 15 Change
On June 15 Anthropic split Agent SDK billing into a separate monthly pool. Plain guide to what changed, what breaks, and how to keep your tools running.
Topic
Step-by-step engineering tutorials with working code. Articles in this section walk through real deploys (ECS, ECR, Fargate, App Runner), real integrations (Claude Code skills, MCP servers, Agent SDK workflows), and real debugging sessions (CannotPullContainerError, Claude Code slowness, MEMORY.md cleanup). Each tutorial is end-to-end: prerequisites, exact commands, expected output, and what to check when something goes wrong. If a guide promises a 10-minute walkthrough, it means 10 minutes of real time, not idealised time.
On June 15 Anthropic split Agent SDK billing into a separate monthly pool. Plain guide to what changed, what breaks, and how to keep your tools running.
What MCP Apps are, how they let MCP servers ship interactive UIs to hosts, the security model, and what to prepare for before the July 28 spec lands.
A plain guide to MCP Apps and the OpenAI Apps SDK in 2026, how they actually relate, the honest differences, code from both sides, and a clear way to pick.
Learn the five deployment strategies (recreate, rolling, blue-green, canary, A/B) with diagrams and pick the right one for your next release without guessing.
How to measure what your Claude Agent SDK workload really costs before the June 15, 2026 credit change, with code, caveats, and a sizing decision.
Microsoft released RAMPART on May 20, 2026, a pytest-native safety testing framework for AI agents. Here is how to apply it to your Claude agents today.
The NSA published MCP security guidance on May 20, 2026. Here are the four named controls, the threats they address, and what to do in your server today.
A practical side by side look at the Claude Agent SDK and Vercel AI SDK 6 for production AI agents in 2026, with code, defaults, and a clear pick.
Claude Code feels slow when context bloats, MCP tools chain in loops, or memory files grow. See the seven fixes with /doctor and /compact walkthroughs.
Seven causes of CannotPullContainerError on ECS Fargate: IAM, networking, image tag, architecture, disk, Docker Hub, VPC endpoints. Match the cause and fix it.
AWS Copilot CLI reaches end of support on June 12, 2026. What AWS recommends migrating to, how to choose, and a service-by-service map.
What Claude Code's /goal command and Agent View do, how they work together, how to write a completion condition that holds, and when to actually use them.
A plain-English breakdown of Anthropic's internal report on how its own teams use Claude Code: the real workflows, the numbers, and what you can copy.
ECS Express Mode takes one container image and builds the load balancer, HTTPS URL, auto scaling, and monitoring. See when to use it vs Fargate or App Runner.
A plain-English guide to Amazon ECR: what it is, the push and pull flow, IAM auth, lifecycle policies, image scanning, pull-through cache, and cost.
A plain-English guide to Amazon ECS: what it is, how it differs from EC2, what ECR does, when to use it, and the full container deployment flow.
Claude Code Ultraplan moves your planning to the cloud while your terminal stays free. Here is what it does, how to use it, and when it actually helps.
AutoDream cleans Claude Code's MEMORY.md between sessions. /dream runs it on demand. See what gets rewritten and how to turn it off.
How Claude Code Outcomes work, why a separate grading agent improves agent quality, and how to write a rubric that catches the failures you actually care about.
What Claude Code Routines are, how scheduled, API, and GitHub triggers work, and seven practical workflows you can copy into your own setup this week.
A plain-English comparison of Claude Code skills, MCP servers, subagents, and hooks. When to pick each, with a decision tree, code, and common confusion.
Build your first Claude Managed Agent step by step: agent, environment, session, streaming events, rate limits, and the production gotchas no one mentions.
How to secure a Model Context Protocol server in production: OAuth 2.1, RFC 8707 audience binding, rate limiting, audit logs, and the mistakes I keep seeing.
Build your first Claude Code skill. See the SKILL.md format, where skills live on disk, the YAML frontmatter fields, and how to test the skill.
What Claude Mythos is, who got access through Project Glasswing, the zero-days it has already found, and what its arrival means for security teams in 2026.
A practical guide to the WhatsApp Liquid Glass design on iOS 26: what changes, how to enable it, and what to do if it has not arrived yet.
A friendly, hands-picked list of the best MCP servers to use in 2026, with notes on what each one does, who it is for, and when it actually helps.
A calm side by side look at the Claude Agent SDK and LangChain for production AI agents, written by someone who has shipped both.
Hooks let Claude Code run small commands before and after it does things. This guide explains what they are, how to set them up, and the few mistakes to avoid.
Learn what agentic RAG is, how it differs from traditional RAG, and how to build one with the Claude Agent SDK and pgvector. Step by step, with full code.
A short and friendly walk through of building a Model Context Protocol server in Node.js. Plain code, no jargon, and the small mistakes to watch for.
A side by side look at Claude Opus 4.7 and GPT-5 in 2026, with notes on coding, reasoning, long context, vision, agent work, speed, and price.