Production AI Agent Systems

Architecture, tools, memory, evaluation, security and deployment for AI Agents that must operate beyond a demo.

Primary Paths

All AI Engineering

Browse the complete reviewed English AI engineering library.

AI Agent Security Infrastructure

Identity, authorization, approval, sandboxing, MCP security, retention, audit and incident response for production agents.

Open →

AI Agent vs AI Assistant

Compare execution autonomy, tool use, state, retries and approval boundaries before choosing an architecture.

Open →

AI Email Routing Agent

Turn shared inbox messages into auditable, SLA-bound tickets with human fallback.

Open →

Related Resources

Context engineering routes system instructions, user tasks, retrieval, memory, tools, state and compacted history into an AI agent to produce grounded answers, completed tasks and lower context cost - XBSTACK

What Is Context Engineering? Reducing AI Agent Cost with Retrieval, Tool Search and Memory

Context engineering goes beyond prompt engineering. This guide combines Microsoft, Anthropic and Goo

2026-09-04

Read Article

10 min

GPT-6 Astra API English cover showing pricing, 1.05M context, 128K output and Responses API migration - XBSTACK

GPT-6 Astra API Guide: Pricing, Claude/Gemini Comparison, 1.05M Context and Migration

GPT-6 Astra API pricing, 1.05M context, Responses migration and rollout, plus Claude Fable 5.1 and G

2026-09-04

Read Article

10 min

Gemini 3.8 Flash versus 3.7 Flash for coding, agents, pricing and per-task token cost - XBSTACK

Gemini 3.8 Flash vs 3.7 Flash: Same Price, Different Agent Cost?

Gemini 3.8 Flash vs 3.7 Flash: compare price, 1M context, thinking levels, AI coding, AIGC, Claude/G

2026-09-03

Read Article

6 min

Claude Fable 5.1 vs Mythos 5.1 comparison of pricing, access, coding, and agent use cases - XBSTACK

Claude Fable 5.1 vs Mythos 5.1: Pricing, Access, Coding, and Agent Trade-offs

Claude Fable 5.1 and Mythos 5.1 share the same underlying model but differ in safeguards and access.

2026-09-02

Read Article

5 min

llms.txt v2 agent discovery with alternate, describedby, and Markdown pages - XBSTACK

How to Configure llms.txt v2: rel=alternate, describedby, and Markdown Discovery

llms.txt v2 adds explicit discovery for Markdown pages and llms.txt through rel=alternate and rel=de

2026-08-31

Read Article

4 min

MCP configuration security review for secrets, shell, remote MCP, and permissions - XBSTACK

How to Security-Check an MCP Configuration: Secrets, Shell, Remote MCP, and Permissions

Review MCP configuration for credential exposure, shell execution, remote MCP trust, permissions, lo

2026-08-31

Read Article

5 min

WebMCP on Chrome 149 with forms, tool schemas, and registerTool - XBSTACK

How to Add WebMCP to a Website: Chrome 149 Forms, Tool Schemas, and registerTool

Add WebMCP to a website with Chrome 149 declarative forms or document.modelContext.registerTool, the

2026-08-31

Read Article

4 min

LangGraph aupdate_state raising Ambiguous update after a START seed while synchronous update_state succeeds - XBSTACK

LangGraph aupdate_state 'Ambiguous update': Why Async Fails When update_state Works

LangGraph 1.2.11 repro: sync update_state works, but async aupdate_state raises Ambiguous update aft

2026-08-26

Read Article

6 min

OpenAI Zero Data Retention architecture for AI agents across Responses API, MCP, memory, audit and data residency - XBSTACK

OpenAI Zero Data Retention for AI Agents: Responses API, MCP and Data Boundaries

OpenAI Zero Data Retention does not make an AI agent stack stateless. This guide maps ZDR across Res

2026-08-20

Read Article

8 min

DeepSeek V4-Pro-0813 and DeepSeek Harness features, pricing, API setup, and agent architecture - XBSTACK

What Is DeepSeek V4-Pro-0813? API Pricing, Setup, and DeepSeek Harness Guide

DeepSeek V4-Pro-0813 explained: API pricing, 1M context, 384K output, concurrency, OpenAI/Anthropic

2026-08-15

Read Article

10 min

Google ADK 2.6.2 state_delta not applied to session.state when Runner.run_async resumes without new_message - XBSTACK

Google ADK Resume Bugs: state_delta Loss and the 2.7.0 A2A HITL Regression

Google ADK state_delta not applied: reproduce the 2.6.2 state-only resume loss and compare the 2.6.1

2026-08-09

Read Article

7 min

OpenAI Responses API streamed function_call breaks at Stream Abort before Conversation commit and returns 400 No tool call found - XBSTACK

OpenAI Responses API: Why Stream Abort Causes No tool call found for function call output

A function_call can be visible before it is durable in Conversation state. This guide explains the 4

2026-08-05

Read Article

13 min

Two OpenAI Agents SDK FunctionTools named lookup are registered while local dispatch keeps only the later tool - XBSTACK

OpenAI Agents SDK Duplicate Tool Names: Why the Later Tool Wins

OpenAI Agents SDK duplicate tool names can trigger a provider 400 or last-wins local dispatch. Repro

2026-08-03

Read Article

9 min

Experiment showing a LangGraph streaming UI ahead of the authoritative SQLite checkpoint after cancellation - XBSTACK

LangGraph State Lost After Cancel: Why Streaming Output Disappears After Stop

LangGraph state lost after cancel or Stop? A LangGraph 1.2.9 experiment shows why streamed output ca

2026-07-30

Read Article

16 min

MCP 2026-07-28 stateless migration with cross-replica session failure and explicit handles - XBSTACK

MCP initialize Removed: How Do You Replace Mcp-Session-Id in a Remote Server?

Why was MCP initialize removed, and where should state live after Mcp-Session-Id disappears? A two-r

2026-07-28

Read Article

22 min

An AI Agent tool call entering a Policy Gate before allow, confirmation, step-up authentication, or block - XBSTACK

AI Agent Tool Authorization: Policy Gates, HITL, and Per-Call Access Control

AI Agent tool authorization: Why can an AI Agent pass guardrails and still execute an unauthorized t

2026-07-27

Read Article

7 min

An n8n AI Agent with connected tools branching to a skipped text response or a real tool execution - XBSTACK

n8n AI Agent Not Calling Tools: tool_choice, Provider Compatibility, and Memory

n8n AI Agent not calling tools: diagnose tool_choice, provider compatibility, tool schema and descri

2026-07-26

Read Article

7 min

Production architecture for resuming an OpenAI Agents SDK RunState Tool Approval across processes - XBSTACK

OpenAI Agents SDK RunState: How to Resume Tool Approval Across Processes

Resume OpenAI Agents SDK Tool Approval with RunState across processes. Test to_json/from_json, appro

2026-07-24

Read Article

18 min

Vercel AI SDK 7 interrupted streams, Cloudflare 524 boundaries, message persistence, and idempotent tool-call recovery - XBSTACK

Vercel AI SDK 7 Migration: Tool Calls, WorkflowAgent Approval, and Stream Recovery

Vercel AI SDK 7 migration with reproducible Tool Call, persistence, abort, and WorkflowAgent approva

2026-07-21

Read Article

25 min

AI Agent memory retrieval architecture with structured lookup, vector recall, re-ranking and prompt assembly - XBSTACK

AI Agent Memory Retrieval Architecture: Hybrid Search, Re-ranking, Freshness and Conflict Resolution

A production-focused guide to AI Agent memory retrieval. Design a safe retrieval pipeline with ident

2026-07-16

Read Article

15 min

Division of labor and complete workflow for ChatGPT Work, Chat, and Codex in a real website task - XBSTACK

ChatGPT Chat vs Work: What’s the Difference? When to Use Codex

ChatGPT Work vs Codex: compare Chat, Work, and Codex by task boundary, then check the latest Cloud W

2026-07-13

Read Article

12 min

GPT-5.6 Tested in Real Projects, Content Creation, and Website Data Analysis - XBSTACK

GPT-5.6 Review: Real Coding, Content and Data Tests

GPT-5.6 review from a real Astro project: coding, content operations, Search Console and GA4 tests,

2026-07-10

Read Article

19 min

Cover image for the Claude Sonnet 5 AI Coding Agent hands-on test optimizing oversized Astro website chunks - XBSTACK

Claude Sonnet 5 Hands-On: Full Process for Optimizing Oversized Astro Chunks

Claude Sonnet 5 Hands-On: This article tests Claude Sonnet 5 on a real XBSTACK Astro project, using

2026-07-08

Read Article

10 min

Diagram of the three-layer architecture of MCP, Function Calling, and API Gateway in AI Agent tool integration - XBSTACK

How to Configure MCP, Function Calling, and API Gateway: A Three-Layer Architecture for AI Agent Tool Integration

MCP, Function Calling, and API Gateway solve different layers of AI tool integration: model tool sel

Mcp

2026-07-07

Read Article

9 min