Notes on agents, software, and work.
Systems / September 7, 2026 / 10 min read
Multi-agent orchestration: Claude Code design guide
Claude Code multi-agent orchestration starts with dependencies, file ownership, permissions, handoffs, and validation before agents begin work.
Systems / September 6, 2026 / 11 min read
LangGraph with Claude Agent SDK: draw the boundary before you integrate
LangGraph can own workflow state and routing while Claude Agent SDK runs the tool loop, with approvals, persistence, and recovery assigned explicitly.
Systems / September 5, 2026 / 10 min read
Human in the loop framework: design the boundary, not just the approval
A human-in-the-loop framework needs explicit authority, review context, decision states, durable holds, recovery, and an audit trail.
Systems / September 4, 2026 / 9 min read
AI agent orchestration example: trace the workflow from routing to recovery
A worked AI agent orchestration example traces routing, approvals, checkpoints, recovery, and the criteria that determine framework fit.
Systems / September 3, 2026 / 10 min read
Best multi agent orchestration framework: choose by workload and control
The best multi-agent orchestration framework is the one that fits your workload's coordination, control, recovery, cost, and maintenance needs.
Systems / September 2, 2026 / 10 min read
Agent orchestration open source: choose the smallest architecture that fits
The smallest viable open-source agent orchestrator fits the workload, authority boundaries, recovery needs, and operating burden.
Systems / September 1, 2026 / 10 min read
Multi-agent systems with LangGraph: choose the architecture by workload and control
Choose a LangGraph multi-agent pattern by workload, call and token costs, state boundaries, approval pauses, and side-effect replay.
Systems / August 31, 2026 / 9 min read
LangGraph vs MCP: choose the right architecture boundary
LangGraph manages stateful agent execution while MCP standardizes external tools, so the same architecture can use both without confusing their responsibilities.
Systems / August 30, 2026 / 11 min read
How to build multi-agent systems without losing control
A multi-agent system earns its complexity when parallel work, separate context, specialized tools, or decision checkpoints justify the coordination cost.