Agent & AI Insights
Claude Code know-how, heterogeneous agents, harness engineering, and more — trends and adoption know-how from the Marblo team
Make Three Models Refute Your Hypothesis — the Marblo Adversarial Verification Board
Ask an AI and it usually says 'great idea.' What you actually need is refutation. Here's a Marblo recipe that makes different-vendor agents attack one hypothesis, design, or decision from different angles — so only the conclusions that survive get adopted.
Plan → Code → Review → Test on One Board — Running a Dev Pipeline in Marblo
Coding got faster but shipping didn't, and the reason is usually the steps around the code and the human handoffs between them. Here's a Marblo dev-pipeline recipe that splits task breakdown, implementation, review, and testing across agents while the team collaborates on one board.
Turning an Inbox Flood Into 3-Agent Triage — Automating First-Line CS With a Marblo Board
Half of your daily tickets are the same repeat questions with fixed answers. Here's a Marblo board recipe that splits classification, drafting, and safety-checking across separate agents — so your reps only handle what needs judgment.
Weekly Marketing Reports, From 30 Minutes to 3 — Automating Them With a Marblo Board
Still copy-pasting GA4, Naver, and Meta numbers into a report every week? Here's a Marblo board recipe that splits collection, analysis, writing, and verification across separate agents — step by step.
Claude Code + MCP in Real Workflows — Notes from a Korean AI Studio
How a Seoul-based AI agency runs Claude Code with MCP servers across every project. The patterns that actually scale, the integrations that paid off, and the workflow tax we eliminated.
Building Your First Multi-Agent System with Marblo — Hands-On Tutorial
Walk through building a real multi-agent workflow in Marblo, from blank board to production deploy in under thirty minutes. Researcher, writer, fact-checker — heterogeneous models, MCP tools, and observability included.
MCP Servers in Production — Authentication, Rate Limits, and Observability
Building MCP (Model Context Protocol) servers for a hobby project is easy. Running them in production with real authentication, real rate limits, and traces you can debug at 2 AM is a different problem. This is what we learned.
Heterogeneous Agents in Production — Why Single-Model Setups Fail at Scale
After running heterogeneous AI agents in production for 18 months, we measured what single-vendor setups give up. The cost premium, the failure modes, and the team-level patterns that only work when you mix models on purpose.
AI Agent Orchestration Platforms in 2026 — LangGraph, CrewAI, AutoGen, and Marblo Compared
An engineering-grade comparison of the major AI agent orchestration platforms in 2026. Where each one shines, where each one breaks, and which choice fits which workload — from prototype to multi-team production.
Codex 5.5 GOAL Mode — The New Standard for Autonomous Agents
OpenAI Codex's latest GOAL mode is reshaping the autonomous agent paradigm. We analyze the shift from imperative commands to goal-driven execution, and how it compares to Marblo's natural language orchestrator.
Why Heterogeneous AI Agents Beat Single-Model — Claude, GPT, and Gemini on One Board
Why leading AI teams in 2026 are choosing heterogeneous agent orchestration over single-vendor solutions. The performance gap and cost efficiency that comes from role-based model assignment — Claude reasoning, GPT generation, Gemini verification.
Model Context Protocol (MCP) Explained — The Standard for Tool-Wielding Agents
MCP gives AI agents standardized access to filesystems, databases, APIs, and Git. We unpack why MCP became the 'USB-C' of the AI agent industry — and how to integrate it with internal company systems.
Claude Code Subagents vs. Real Multi-Agent Orchestration — What's the Difference?
We dissect the gap between Claude Code's 'subagent' pattern and genuine heterogeneous multi-agent orchestration. Single-model N agents vs. heterogeneous N agents, CLI vs. kanban board, shared context vs. physical isolation.
5 Principles for In-house AI Agent Governance — Design the Trust Hierarchy First
When you adopt AI agents internally, the first question is 'Who is responsible for code an agent wrote?' This article lays out 5 governance principles: privilege separation, audit logs, rollback paths, measurable KPIs, and team training.